path
stringlengths 13
17
| screenshot_names
sequencelengths 1
873
| code
stringlengths 0
40.4k
| cell_type
stringclasses 1
value |
|---|---|---|---|
34142232/cell_13 | [
"text_plain_output_1.png",
"image_output_1.png"
] | import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/pm25-mean-annual-exposure/PM25_MAE.csv', index_col=0)
df = df.drop(['Country Code', 'Indicator Name', 'Indicator Code'], axis=1)
df = df.dropna(thresh=10, axis=1)
before = df.shape[0]
na_free = df.dropna(thresh=10, axis=0)
only_na = df[~df.index.isin(na_free.index)]
after = na_free.shape[0]
namelist = only_na.index
df = na_free.transpose()
import matplotlib.pyplot as plt
import datetime as dt
latest = df.tail(1).transpose()
latest | code |
34142232/cell_9 | [
"text_plain_output_1.png",
"image_output_1.png"
] | import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/pm25-mean-annual-exposure/PM25_MAE.csv', index_col=0)
df = df.drop(['Country Code', 'Indicator Name', 'Indicator Code'], axis=1)
df = df.dropna(thresh=10, axis=1)
before = df.shape[0]
na_free = df.dropna(thresh=10, axis=0)
only_na = df[~df.index.isin(na_free.index)]
after = na_free.shape[0]
namelist = only_na.index
df = na_free.transpose()
import matplotlib.pyplot as plt
import datetime as dt
df.plot(y='World') | code |
34142232/cell_6 | [
"text_html_output_1.png"
] | import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/pm25-mean-annual-exposure/PM25_MAE.csv', index_col=0)
df = df.drop(['Country Code', 'Indicator Name', 'Indicator Code'], axis=1)
df = df.dropna(thresh=10, axis=1)
before = df.shape[0]
na_free = df.dropna(thresh=10, axis=0)
only_na = df[~df.index.isin(na_free.index)]
after = na_free.shape[0]
print(str(before - after) + " countries don't have any PM2.5 data reported:")
namelist = only_na.index
for name in namelist:
print(name) | code |
34142232/cell_11 | [
"text_plain_output_1.png"
] | import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/pm25-mean-annual-exposure/PM25_MAE.csv', index_col=0)
df = df.drop(['Country Code', 'Indicator Name', 'Indicator Code'], axis=1)
df = df.dropna(thresh=10, axis=1)
before = df.shape[0]
na_free = df.dropna(thresh=10, axis=0)
only_na = df[~df.index.isin(na_free.index)]
after = na_free.shape[0]
namelist = only_na.index
df = na_free.transpose()
import matplotlib.pyplot as plt
import datetime as dt
df[['World', 'China']].plot(kind='bar') | code |
34142232/cell_1 | [
"text_plain_output_1.png"
] | import os
import numpy as np
import pandas as pd
import os
for dirname, _, filenames in os.walk('/kaggle/input'):
for filename in filenames:
print(os.path.join(dirname, filename)) | code |
34142232/cell_14 | [
"text_plain_output_1.png",
"image_output_1.png"
] | import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/pm25-mean-annual-exposure/PM25_MAE.csv', index_col=0)
df = df.drop(['Country Code', 'Indicator Name', 'Indicator Code'], axis=1)
df = df.dropna(thresh=10, axis=1)
before = df.shape[0]
na_free = df.dropna(thresh=10, axis=0)
only_na = df[~df.index.isin(na_free.index)]
after = na_free.shape[0]
namelist = only_na.index
df = na_free.transpose()
import matplotlib.pyplot as plt
import datetime as dt
#fetch the latest data for all countries
latest = df.tail(1).transpose()
latest
latest['2017'].plot(kind='pie') | code |
2036047/cell_9 | [
"text_plain_output_1.png"
] | import pandas as pd
import plotly.graph_objs as go
import plotly.offline as pyo
data_df = pd.read_csv('../input/mushrooms.csv')
data_df['y'] = data_df['class'].map({'p': 1, 'e': 0})
feature_columns = [c for c in data_df.columns if not c in ('class', 'y')]
stats_df = []
single_val_c = {}
for i, c in enumerate(feature_columns):
if data_df[c].nunique() == 1:
single_val_c[c] = data_df[c].unique()[0]
continue
gb = data_df.groupby(c)
m = gb['y'].mean()
s = gb.size()
df = pd.DataFrame(index=range(len(m)))
df['col'] = c
df['val'] = m.index.values
df['positive_percentage'] = m.values
df['quantity_percentage'] = s.values / s.sum()
stats_df.append(df)
trace_prate = go.Bar(x=df['val'], y=df['positive_percentage'] * 100, name='positive percentage')
trace_cnt = go.Bar(x=df['val'], y=df['quantity_percentage'] * 100, name='quantity percentage')
layout = go.Layout(xaxis=dict(title=c), yaxis=dict(title='positive and quantity percentage'))
fig = go.Figure(data=[trace_prate, trace_cnt], layout=layout)
stats_df = pd.concat(stats_df, axis=0)
for c in single_val_c.keys():
print('The column %s only has one unique value with %r.' % (c, single_val_c[c]))
print('It does work for the classification, which will be removed.')
feature_columns.remove(c)
data_df = data_df[feature_columns + ['y']].copy() | code |
2036047/cell_25 | [
"text_plain_output_1.png"
] | from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier
from sklearn.linear_model import RidgeClassifier, LogisticRegressionCV
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
from sklearn.model_selection import GridSearchCV, train_test_split
from sklearn.naive_bayes import GaussianNB, MultinomialNB
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import LabelEncoder, PolynomialFeatures
from time import time
import pandas as pd
import plotly.graph_objs as go
import plotly.offline as pyo
data_df = pd.read_csv('../input/mushrooms.csv')
data_df['y'] = data_df['class'].map({'p': 1, 'e': 0})
feature_columns = [c for c in data_df.columns if not c in ('class', 'y')]
stats_df = []
single_val_c = {}
for i, c in enumerate(feature_columns):
if data_df[c].nunique() == 1:
single_val_c[c] = data_df[c].unique()[0]
continue
gb = data_df.groupby(c)
m = gb['y'].mean()
s = gb.size()
df = pd.DataFrame(index=range(len(m)))
df['col'] = c
df['val'] = m.index.values
df['positive_percentage'] = m.values
df['quantity_percentage'] = s.values / s.sum()
stats_df.append(df)
trace_prate = go.Bar(x=df['val'], y=df['positive_percentage'] * 100, name='positive percentage')
trace_cnt = go.Bar(x=df['val'], y=df['quantity_percentage'] * 100, name='quantity percentage')
layout = go.Layout(xaxis=dict(title=c), yaxis=dict(title='positive and quantity percentage'))
fig = go.Figure(data=[trace_prate, trace_cnt], layout=layout)
stats_df = pd.concat(stats_df, axis=0)
for c in single_val_c.keys():
feature_columns.remove(c)
data_df = data_df[feature_columns + ['y']].copy()
data_df = pd.concat((data_df, data_df), axis=0, ignore_index=True)
data_all = pd.get_dummies(data=data_df, columns=feature_columns, prefix=feature_columns)
def grid_search(base_model, param_grid, X_train, y_train):
gs_c = GridSearchCV(base_model, param_grid=param_grid, n_jobs=-1, cv=3)
gs_c.fit(X_train, y_train)
return gs_c
def ridge_model(X_train, y_train):
r_c = Pipeline([('poly', PolynomialFeatures(interaction_only=True)), ('clf', RidgeClassifier(random_state=1))])
params_pool = dict(poly__degree=[2], clf__alpha=[0.01, 0.03, 0.1, 0.3, 1])
return grid_search(r_c, params_pool, X_train, y_train)
def randomForest_model(X_train, y_train):
rf_c = RandomForestClassifier(random_state=1)
params_pool = dict(max_depth=[5, 7, 9], max_features=[0.3, 0.5], n_estimators=[12, 20, 36, 50])
return grid_search(rf_c, params_pool, X_train, y_train)
def gaussianNB_model(X_train, y_train):
gnb = Pipeline([('poly', PolynomialFeatures(interaction_only=True)), ('clf', GaussianNB())])
gnb.fit(X_train, y_train)
return gnb
def multinomialNB_model(X_train, y_train):
mnb = Pipeline([('poly', PolynomialFeatures(interaction_only=True)), ('clf', MultinomialNB(alpha=1e-05))])
mnb.fit(X_train, y_train)
return mnb
def do_model_train(model_name, X_train, y_train, X_test, y_test):
bg = time()
if 'Ridge' == model_name:
model = ridge_model(X_train, y_train)
elif 'RandomForest' == model_name:
model = randomForest_model(X_train, y_train)
elif 'GaussianNB' == model_name:
model = gaussianNB_model(X_train, y_train)
elif 'multinomialNB' == model_name:
model = multinomialNB_model(X_train, y_train)
y_hat = model.predict(X_train)
y_hat = model.predict(X_test)
def clean_data_df(df, threshold=0.02):
feature_convert = dict()
for col, sub in stats_df.groupby('col'):
ns = sub[sub.quantity_percentage < threshold]
n_ns = sub[sub.quantity_percentage >= threshold]
for idx in ns.index:
if ns.loc[idx, 'positive_percentage'] > 0.5:
p_n_ns = n_ns[n_ns.positive_percentage > 0.5]
if not p_n_ns.empty:
feature_convert.setdefault(col, []).append((ns.loc[idx, 'val'], p_n_ns['val'].values[0]))
df.loc[df[col] == ns.loc[idx, 'val'], col] = p_n_ns['val'].values[0]
else:
n_n_ns = n_ns[n_ns.positive_percentage <= 0.5]
if not n_n_ns.empty:
feature_convert.setdefault(col, []).append((ns.loc[idx, 'val'], n_n_ns['val'].values[0]))
df.loc[df[col] == ns.loc[idx, 'val'], col] = n_n_ns['val'].values[0]
return (pd.get_dummies(data=df, columns=feature_columns, prefix=feature_columns), feature_convert)
cleaned_df, feature_convert = clean_data_df(data_df.copy())
X_all, y_all = (cleaned_df[[c for c in cleaned_df.columns if c != 'y']], cleaned_df['y'])
X_train, X_test, y_train, y_test = train_test_split(X_all, y_all, test_size=0.33, random_state=1)
do_model_train('RandomForest', X_train, y_train, X_test, y_test) | code |
2036047/cell_4 | [
"text_plain_output_1.png"
] | import pandas as pd
data_df = pd.read_csv('../input/mushrooms.csv')
data_df.info() | code |
2036047/cell_23 | [
"text_html_output_1.png"
] | import pandas as pd
import plotly.graph_objs as go
import plotly.offline as pyo
data_df = pd.read_csv('../input/mushrooms.csv')
data_df['y'] = data_df['class'].map({'p': 1, 'e': 0})
feature_columns = [c for c in data_df.columns if not c in ('class', 'y')]
stats_df = []
single_val_c = {}
for i, c in enumerate(feature_columns):
if data_df[c].nunique() == 1:
single_val_c[c] = data_df[c].unique()[0]
continue
gb = data_df.groupby(c)
m = gb['y'].mean()
s = gb.size()
df = pd.DataFrame(index=range(len(m)))
df['col'] = c
df['val'] = m.index.values
df['positive_percentage'] = m.values
df['quantity_percentage'] = s.values / s.sum()
stats_df.append(df)
trace_prate = go.Bar(x=df['val'], y=df['positive_percentage'] * 100, name='positive percentage')
trace_cnt = go.Bar(x=df['val'], y=df['quantity_percentage'] * 100, name='quantity percentage')
layout = go.Layout(xaxis=dict(title=c), yaxis=dict(title='positive and quantity percentage'))
fig = go.Figure(data=[trace_prate, trace_cnt], layout=layout)
stats_df = pd.concat(stats_df, axis=0)
for c in single_val_c.keys():
feature_columns.remove(c)
data_df = data_df[feature_columns + ['y']].copy()
data_df = pd.concat((data_df, data_df), axis=0, ignore_index=True)
data_all = pd.get_dummies(data=data_df, columns=feature_columns, prefix=feature_columns)
def clean_data_df(df, threshold=0.02):
feature_convert = dict()
for col, sub in stats_df.groupby('col'):
ns = sub[sub.quantity_percentage < threshold]
n_ns = sub[sub.quantity_percentage >= threshold]
for idx in ns.index:
if ns.loc[idx, 'positive_percentage'] > 0.5:
p_n_ns = n_ns[n_ns.positive_percentage > 0.5]
if not p_n_ns.empty:
feature_convert.setdefault(col, []).append((ns.loc[idx, 'val'], p_n_ns['val'].values[0]))
df.loc[df[col] == ns.loc[idx, 'val'], col] = p_n_ns['val'].values[0]
else:
n_n_ns = n_ns[n_ns.positive_percentage <= 0.5]
if not n_n_ns.empty:
feature_convert.setdefault(col, []).append((ns.loc[idx, 'val'], n_n_ns['val'].values[0]))
df.loc[df[col] == ns.loc[idx, 'val'], col] = n_n_ns['val'].values[0]
return (pd.get_dummies(data=df, columns=feature_columns, prefix=feature_columns), feature_convert)
cleaned_df, feature_convert = clean_data_df(data_df.copy())
cleaned_df.info() | code |
2036047/cell_20 | [
"text_html_output_10.png",
"text_html_output_16.png",
"text_html_output_4.png",
"text_html_output_6.png",
"text_html_output_2.png",
"text_html_output_15.png",
"text_html_output_5.png",
"text_html_output_14.png",
"text_html_output_19.png",
"text_html_output_9.png",
"text_html_output_13.png",
"text_html_output_20.png",
"text_html_output_21.png",
"text_html_output_1.png",
"text_html_output_17.png",
"text_html_output_18.png",
"text_html_output_12.png",
"text_html_output_11.png",
"text_html_output_8.png",
"text_html_output_3.png",
"text_html_output_7.png"
] | from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier
from sklearn.linear_model import RidgeClassifier, LogisticRegressionCV
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
from sklearn.model_selection import GridSearchCV, train_test_split
from sklearn.naive_bayes import GaussianNB, MultinomialNB
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import LabelEncoder, PolynomialFeatures
from time import time
import pandas as pd
import plotly.graph_objs as go
import plotly.offline as pyo
data_df = pd.read_csv('../input/mushrooms.csv')
data_df['y'] = data_df['class'].map({'p': 1, 'e': 0})
feature_columns = [c for c in data_df.columns if not c in ('class', 'y')]
stats_df = []
single_val_c = {}
for i, c in enumerate(feature_columns):
if data_df[c].nunique() == 1:
single_val_c[c] = data_df[c].unique()[0]
continue
gb = data_df.groupby(c)
m = gb['y'].mean()
s = gb.size()
df = pd.DataFrame(index=range(len(m)))
df['col'] = c
df['val'] = m.index.values
df['positive_percentage'] = m.values
df['quantity_percentage'] = s.values / s.sum()
stats_df.append(df)
trace_prate = go.Bar(x=df['val'], y=df['positive_percentage'] * 100, name='positive percentage')
trace_cnt = go.Bar(x=df['val'], y=df['quantity_percentage'] * 100, name='quantity percentage')
layout = go.Layout(xaxis=dict(title=c), yaxis=dict(title='positive and quantity percentage'))
fig = go.Figure(data=[trace_prate, trace_cnt], layout=layout)
stats_df = pd.concat(stats_df, axis=0)
for c in single_val_c.keys():
feature_columns.remove(c)
data_df = data_df[feature_columns + ['y']].copy()
data_df = pd.concat((data_df, data_df), axis=0, ignore_index=True)
data_all = pd.get_dummies(data=data_df, columns=feature_columns, prefix=feature_columns)
def grid_search(base_model, param_grid, X_train, y_train):
gs_c = GridSearchCV(base_model, param_grid=param_grid, n_jobs=-1, cv=3)
gs_c.fit(X_train, y_train)
return gs_c
def ridge_model(X_train, y_train):
r_c = Pipeline([('poly', PolynomialFeatures(interaction_only=True)), ('clf', RidgeClassifier(random_state=1))])
params_pool = dict(poly__degree=[2], clf__alpha=[0.01, 0.03, 0.1, 0.3, 1])
return grid_search(r_c, params_pool, X_train, y_train)
def randomForest_model(X_train, y_train):
rf_c = RandomForestClassifier(random_state=1)
params_pool = dict(max_depth=[5, 7, 9], max_features=[0.3, 0.5], n_estimators=[12, 20, 36, 50])
return grid_search(rf_c, params_pool, X_train, y_train)
def gaussianNB_model(X_train, y_train):
gnb = Pipeline([('poly', PolynomialFeatures(interaction_only=True)), ('clf', GaussianNB())])
gnb.fit(X_train, y_train)
return gnb
def multinomialNB_model(X_train, y_train):
mnb = Pipeline([('poly', PolynomialFeatures(interaction_only=True)), ('clf', MultinomialNB(alpha=1e-05))])
mnb.fit(X_train, y_train)
return mnb
def do_model_train(model_name, X_train, y_train, X_test, y_test):
bg = time()
if 'Ridge' == model_name:
model = ridge_model(X_train, y_train)
elif 'RandomForest' == model_name:
model = randomForest_model(X_train, y_train)
elif 'GaussianNB' == model_name:
model = gaussianNB_model(X_train, y_train)
elif 'multinomialNB' == model_name:
model = multinomialNB_model(X_train, y_train)
y_hat = model.predict(X_train)
y_hat = model.predict(X_test)
X_all, y_all = (data_all[[c for c in data_all.columns if c != 'y']], data_all['y'])
X_train, X_test, y_train, y_test = train_test_split(X_all, y_all, test_size=0.33, random_state=1)
do_model_train('RandomForest', X_train, y_train, X_test, y_test) | code |
2036047/cell_7 | [
"text_plain_output_1.png"
] | import pandas as pd
import plotly.graph_objs as go
import plotly.offline as pyo
data_df = pd.read_csv('../input/mushrooms.csv')
data_df['y'] = data_df['class'].map({'p': 1, 'e': 0})
feature_columns = [c for c in data_df.columns if not c in ('class', 'y')]
stats_df = []
single_val_c = {}
for i, c in enumerate(feature_columns):
if data_df[c].nunique() == 1:
single_val_c[c] = data_df[c].unique()[0]
continue
gb = data_df.groupby(c)
m = gb['y'].mean()
s = gb.size()
df = pd.DataFrame(index=range(len(m)))
df['col'] = c
df['val'] = m.index.values
df['positive_percentage'] = m.values
df['quantity_percentage'] = s.values / s.sum()
stats_df.append(df)
trace_prate = go.Bar(x=df['val'], y=df['positive_percentage'] * 100, name='positive percentage')
trace_cnt = go.Bar(x=df['val'], y=df['quantity_percentage'] * 100, name='quantity percentage')
layout = go.Layout(xaxis=dict(title=c), yaxis=dict(title='positive and quantity percentage'))
fig = go.Figure(data=[trace_prate, trace_cnt], layout=layout)
pyo.iplot(fig)
stats_df = pd.concat(stats_df, axis=0) | code |
2036047/cell_8 | [
"text_html_output_1.png",
"text_plain_output_1.png"
] | import pandas as pd
import plotly.graph_objs as go
import plotly.offline as pyo
data_df = pd.read_csv('../input/mushrooms.csv')
data_df['y'] = data_df['class'].map({'p': 1, 'e': 0})
feature_columns = [c for c in data_df.columns if not c in ('class', 'y')]
stats_df = []
single_val_c = {}
for i, c in enumerate(feature_columns):
if data_df[c].nunique() == 1:
single_val_c[c] = data_df[c].unique()[0]
continue
gb = data_df.groupby(c)
m = gb['y'].mean()
s = gb.size()
df = pd.DataFrame(index=range(len(m)))
df['col'] = c
df['val'] = m.index.values
df['positive_percentage'] = m.values
df['quantity_percentage'] = s.values / s.sum()
stats_df.append(df)
trace_prate = go.Bar(x=df['val'], y=df['positive_percentage'] * 100, name='positive percentage')
trace_cnt = go.Bar(x=df['val'], y=df['quantity_percentage'] * 100, name='quantity percentage')
layout = go.Layout(xaxis=dict(title=c), yaxis=dict(title='positive and quantity percentage'))
fig = go.Figure(data=[trace_prate, trace_cnt], layout=layout)
stats_df = pd.concat(stats_df, axis=0)
stats_df.describe() | code |
2036047/cell_16 | [
"text_plain_output_1.png"
] | import pandas as pd
import plotly.graph_objs as go
import plotly.offline as pyo
data_df = pd.read_csv('../input/mushrooms.csv')
data_df['y'] = data_df['class'].map({'p': 1, 'e': 0})
feature_columns = [c for c in data_df.columns if not c in ('class', 'y')]
stats_df = []
single_val_c = {}
for i, c in enumerate(feature_columns):
if data_df[c].nunique() == 1:
single_val_c[c] = data_df[c].unique()[0]
continue
gb = data_df.groupby(c)
m = gb['y'].mean()
s = gb.size()
df = pd.DataFrame(index=range(len(m)))
df['col'] = c
df['val'] = m.index.values
df['positive_percentage'] = m.values
df['quantity_percentage'] = s.values / s.sum()
stats_df.append(df)
trace_prate = go.Bar(x=df['val'], y=df['positive_percentage'] * 100, name='positive percentage')
trace_cnt = go.Bar(x=df['val'], y=df['quantity_percentage'] * 100, name='quantity percentage')
layout = go.Layout(xaxis=dict(title=c), yaxis=dict(title='positive and quantity percentage'))
fig = go.Figure(data=[trace_prate, trace_cnt], layout=layout)
stats_df = pd.concat(stats_df, axis=0)
for c in single_val_c.keys():
feature_columns.remove(c)
data_df = data_df[feature_columns + ['y']].copy()
data_df = pd.concat((data_df, data_df), axis=0, ignore_index=True)
data_all = pd.get_dummies(data=data_df, columns=feature_columns, prefix=feature_columns)
data_all.info() | code |
2036047/cell_3 | [
"text_plain_output_1.png"
] | from subprocess import check_output
np.set_printoptions(suppress=True, linewidth=300)
pd.options.display.float_format = lambda x: '%0.6f' % x
pyo.init_notebook_mode(connected=True)
print(check_output(['ls', '../input']).decode('utf-8')) | code |
2036047/cell_17 | [
"text_html_output_1.png"
] | import pandas as pd
import plotly.graph_objs as go
import plotly.offline as pyo
data_df = pd.read_csv('../input/mushrooms.csv')
data_df['y'] = data_df['class'].map({'p': 1, 'e': 0})
feature_columns = [c for c in data_df.columns if not c in ('class', 'y')]
stats_df = []
single_val_c = {}
for i, c in enumerate(feature_columns):
if data_df[c].nunique() == 1:
single_val_c[c] = data_df[c].unique()[0]
continue
gb = data_df.groupby(c)
m = gb['y'].mean()
s = gb.size()
df = pd.DataFrame(index=range(len(m)))
df['col'] = c
df['val'] = m.index.values
df['positive_percentage'] = m.values
df['quantity_percentage'] = s.values / s.sum()
stats_df.append(df)
trace_prate = go.Bar(x=df['val'], y=df['positive_percentage'] * 100, name='positive percentage')
trace_cnt = go.Bar(x=df['val'], y=df['quantity_percentage'] * 100, name='quantity percentage')
layout = go.Layout(xaxis=dict(title=c), yaxis=dict(title='positive and quantity percentage'))
fig = go.Figure(data=[trace_prate, trace_cnt], layout=layout)
stats_df = pd.concat(stats_df, axis=0)
for c in single_val_c.keys():
feature_columns.remove(c)
data_df = data_df[feature_columns + ['y']].copy()
data_df = pd.concat((data_df, data_df), axis=0, ignore_index=True)
data_all = pd.get_dummies(data=data_df, columns=feature_columns, prefix=feature_columns)
data_all.head() | code |
2036047/cell_24 | [
"text_plain_output_1.png"
] | from pprint import pprint
import pandas as pd
import plotly.graph_objs as go
import plotly.offline as pyo
data_df = pd.read_csv('../input/mushrooms.csv')
data_df['y'] = data_df['class'].map({'p': 1, 'e': 0})
feature_columns = [c for c in data_df.columns if not c in ('class', 'y')]
stats_df = []
single_val_c = {}
for i, c in enumerate(feature_columns):
if data_df[c].nunique() == 1:
single_val_c[c] = data_df[c].unique()[0]
continue
gb = data_df.groupby(c)
m = gb['y'].mean()
s = gb.size()
df = pd.DataFrame(index=range(len(m)))
df['col'] = c
df['val'] = m.index.values
df['positive_percentage'] = m.values
df['quantity_percentage'] = s.values / s.sum()
stats_df.append(df)
trace_prate = go.Bar(x=df['val'], y=df['positive_percentage'] * 100, name='positive percentage')
trace_cnt = go.Bar(x=df['val'], y=df['quantity_percentage'] * 100, name='quantity percentage')
layout = go.Layout(xaxis=dict(title=c), yaxis=dict(title='positive and quantity percentage'))
fig = go.Figure(data=[trace_prate, trace_cnt], layout=layout)
stats_df = pd.concat(stats_df, axis=0)
for c in single_val_c.keys():
feature_columns.remove(c)
data_df = data_df[feature_columns + ['y']].copy()
data_df = pd.concat((data_df, data_df), axis=0, ignore_index=True)
data_all = pd.get_dummies(data=data_df, columns=feature_columns, prefix=feature_columns)
def clean_data_df(df, threshold=0.02):
feature_convert = dict()
for col, sub in stats_df.groupby('col'):
ns = sub[sub.quantity_percentage < threshold]
n_ns = sub[sub.quantity_percentage >= threshold]
for idx in ns.index:
if ns.loc[idx, 'positive_percentage'] > 0.5:
p_n_ns = n_ns[n_ns.positive_percentage > 0.5]
if not p_n_ns.empty:
feature_convert.setdefault(col, []).append((ns.loc[idx, 'val'], p_n_ns['val'].values[0]))
df.loc[df[col] == ns.loc[idx, 'val'], col] = p_n_ns['val'].values[0]
else:
n_n_ns = n_ns[n_ns.positive_percentage <= 0.5]
if not n_n_ns.empty:
feature_convert.setdefault(col, []).append((ns.loc[idx, 'val'], n_n_ns['val'].values[0]))
df.loc[df[col] == ns.loc[idx, 'val'], col] = n_n_ns['val'].values[0]
return (pd.get_dummies(data=df, columns=feature_columns, prefix=feature_columns), feature_convert)
cleaned_df, feature_convert = clean_data_df(data_df.copy())
pprint(feature_convert) | code |
2036047/cell_14 | [
"text_plain_output_1.png"
] | import pandas as pd
import plotly.graph_objs as go
import plotly.offline as pyo
data_df = pd.read_csv('../input/mushrooms.csv')
data_df['y'] = data_df['class'].map({'p': 1, 'e': 0})
feature_columns = [c for c in data_df.columns if not c in ('class', 'y')]
stats_df = []
single_val_c = {}
for i, c in enumerate(feature_columns):
if data_df[c].nunique() == 1:
single_val_c[c] = data_df[c].unique()[0]
continue
gb = data_df.groupby(c)
m = gb['y'].mean()
s = gb.size()
df = pd.DataFrame(index=range(len(m)))
df['col'] = c
df['val'] = m.index.values
df['positive_percentage'] = m.values
df['quantity_percentage'] = s.values / s.sum()
stats_df.append(df)
trace_prate = go.Bar(x=df['val'], y=df['positive_percentage'] * 100, name='positive percentage')
trace_cnt = go.Bar(x=df['val'], y=df['quantity_percentage'] * 100, name='quantity percentage')
layout = go.Layout(xaxis=dict(title=c), yaxis=dict(title='positive and quantity percentage'))
fig = go.Figure(data=[trace_prate, trace_cnt], layout=layout)
stats_df = pd.concat(stats_df, axis=0)
for c in single_val_c.keys():
feature_columns.remove(c)
data_df = data_df[feature_columns + ['y']].copy()
data_df = pd.concat((data_df, data_df), axis=0, ignore_index=True)
data_df.info() | code |
2036047/cell_27 | [
"text_plain_output_1.png"
] | from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier
from sklearn.linear_model import RidgeClassifier, LogisticRegressionCV
from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score
from sklearn.model_selection import GridSearchCV, train_test_split
from sklearn.naive_bayes import GaussianNB, MultinomialNB
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import LabelEncoder, PolynomialFeatures
from time import time
import numpy as np
import pandas as pd
import plotly.graph_objs as go
import plotly.offline as pyo
data_df = pd.read_csv('../input/mushrooms.csv')
data_df['y'] = data_df['class'].map({'p': 1, 'e': 0})
feature_columns = [c for c in data_df.columns if not c in ('class', 'y')]
stats_df = []
single_val_c = {}
for i, c in enumerate(feature_columns):
if data_df[c].nunique() == 1:
single_val_c[c] = data_df[c].unique()[0]
continue
gb = data_df.groupby(c)
m = gb['y'].mean()
s = gb.size()
df = pd.DataFrame(index=range(len(m)))
df['col'] = c
df['val'] = m.index.values
df['positive_percentage'] = m.values
df['quantity_percentage'] = s.values / s.sum()
stats_df.append(df)
trace_prate = go.Bar(x=df['val'], y=df['positive_percentage'] * 100, name='positive percentage')
trace_cnt = go.Bar(x=df['val'], y=df['quantity_percentage'] * 100, name='quantity percentage')
layout = go.Layout(xaxis=dict(title=c), yaxis=dict(title='positive and quantity percentage'))
fig = go.Figure(data=[trace_prate, trace_cnt], layout=layout)
stats_df = pd.concat(stats_df, axis=0)
for c in single_val_c.keys():
feature_columns.remove(c)
data_df = data_df[feature_columns + ['y']].copy()
data_df = pd.concat((data_df, data_df), axis=0, ignore_index=True)
data_all = pd.get_dummies(data=data_df, columns=feature_columns, prefix=feature_columns)
def grid_search(base_model, param_grid, X_train, y_train):
gs_c = GridSearchCV(base_model, param_grid=param_grid, n_jobs=-1, cv=3)
gs_c.fit(X_train, y_train)
return gs_c
def ridge_model(X_train, y_train):
r_c = Pipeline([('poly', PolynomialFeatures(interaction_only=True)), ('clf', RidgeClassifier(random_state=1))])
params_pool = dict(poly__degree=[2], clf__alpha=[0.01, 0.03, 0.1, 0.3, 1])
return grid_search(r_c, params_pool, X_train, y_train)
def randomForest_model(X_train, y_train):
rf_c = RandomForestClassifier(random_state=1)
params_pool = dict(max_depth=[5, 7, 9], max_features=[0.3, 0.5], n_estimators=[12, 20, 36, 50])
return grid_search(rf_c, params_pool, X_train, y_train)
def gaussianNB_model(X_train, y_train):
gnb = Pipeline([('poly', PolynomialFeatures(interaction_only=True)), ('clf', GaussianNB())])
gnb.fit(X_train, y_train)
return gnb
def multinomialNB_model(X_train, y_train):
mnb = Pipeline([('poly', PolynomialFeatures(interaction_only=True)), ('clf', MultinomialNB(alpha=1e-05))])
mnb.fit(X_train, y_train)
return mnb
def do_model_train(model_name, X_train, y_train, X_test, y_test):
bg = time()
if 'Ridge' == model_name:
model = ridge_model(X_train, y_train)
elif 'RandomForest' == model_name:
model = randomForest_model(X_train, y_train)
elif 'GaussianNB' == model_name:
model = gaussianNB_model(X_train, y_train)
elif 'multinomialNB' == model_name:
model = multinomialNB_model(X_train, y_train)
y_hat = model.predict(X_train)
y_hat = model.predict(X_test)
X_all, y_all = (data_all[[c for c in data_all.columns if c != 'y']], data_all['y'])
X_train, X_test, y_train, y_test = train_test_split(X_all, y_all, test_size=0.33, random_state=1)
do_model_train('RandomForest', X_train, y_train, X_test, y_test)
def clean_data_df(df, threshold=0.02):
feature_convert = dict()
for col, sub in stats_df.groupby('col'):
ns = sub[sub.quantity_percentage < threshold]
n_ns = sub[sub.quantity_percentage >= threshold]
for idx in ns.index:
if ns.loc[idx, 'positive_percentage'] > 0.5:
p_n_ns = n_ns[n_ns.positive_percentage > 0.5]
if not p_n_ns.empty:
feature_convert.setdefault(col, []).append((ns.loc[idx, 'val'], p_n_ns['val'].values[0]))
df.loc[df[col] == ns.loc[idx, 'val'], col] = p_n_ns['val'].values[0]
else:
n_n_ns = n_ns[n_ns.positive_percentage <= 0.5]
if not n_n_ns.empty:
feature_convert.setdefault(col, []).append((ns.loc[idx, 'val'], n_n_ns['val'].values[0]))
df.loc[df[col] == ns.loc[idx, 'val'], col] = n_n_ns['val'].values[0]
return (pd.get_dummies(data=df, columns=feature_columns, prefix=feature_columns), feature_convert)
cleaned_df, feature_convert = clean_data_df(data_df.copy())
def get_features_importance(x, y):
rf = RandomForestClassifier(n_estimators=500, class_weight={0: 1, 1: 1 / np.sqrt(np.mean(y))}, max_features=0.75, n_jobs=-1, random_state=1)
rf.fit(x, y)
feature_importance = pd.DataFrame(data={'columns': x.columns, 'importance': rf.feature_importances_})
feature_importance.sort_values(by='importance', axis=0, ascending=False, inplace=True)
feature_importance.loc[:, 'cum_importance'] = feature_importance.importance.cumsum()
return feature_importance
def get_features_corr(df, ycol):
corr_y = df.corr()[ycol].map(np.abs)
corr_y = corr_y[[c for c in corr_y.index if c != ycol]]
corr_y = corr_y / corr_y.sum()
feature_importance = pd.DataFrame(data={'columns': corr_y.index.values, 'importance': corr_y.values})
feature_importance.sort_values(by='importance', axis=0, ascending=False, inplace=True)
feature_importance.loc[:, 'cum_importance'] = feature_importance.importance.cumsum()
return feature_importance
data_all = pd.get_dummies(data=data_df, columns=feature_columns, prefix=feature_columns)
X_all, y_all = (data_all[[c for c in data_all.columns if c != 'y']], data_all['y'])
fi = get_features_importance(X_all, y_all)
bg = time()
accuracyScores, precisionScores, recallScores, f1Scores = ([], [], [], [])
for i in range(len(fi)):
cols = fi.iloc[:i + 1]['columns'].values
model = Pipeline([('poly', PolynomialFeatures(interaction_only=True, degree=2)), ('clf', GaussianNB())])
model.fit(X_all[cols], y_all)
y_p = model.predict(X_all[cols])
accuracyScores.append(accuracy_score(y_true=y_all, y_pred=y_p))
precisionScores.append(precision_score(y_true=y_all, y_pred=y_p))
recallScores.append(recall_score(y_true=y_all, y_pred=y_p))
f1Scores.append(f1_score(y_true=y_all, y_pred=y_p))
if accuracyScores[-1] == 1:
break
print()
print('It took %.3f seconds.' % (time() - bg))
traces = [go.Scatter(x=np.arange(len(fi)) + 1, y=fi['cum_importance'][:i + 1], mode='markers+lines', name='importance', text=fi['columns']), go.Scatter(x=np.arange(len(fi)) + 1, y=accuracyScores, mode='markers+lines', name='accuracy Score'), go.Scatter(x=np.arange(len(fi)) + 1, y=precisionScores, mode='markers+lines', name='precision Score'), go.Scatter(x=np.arange(len(fi)) + 1, y=recallScores, mode='markers+lines', name='recall Score'), go.Scatter(x=np.arange(len(fi)) + 1, y=f1Scores, mode='markers+lines', name='F1 Score')]
layout = go.Layout(title='Feature importance/accuracy/precision/recall/F1 Score on different number of features', xaxis=dict(title='number of features'), yaxis=dict(title='importance/accuracy/precision/recall/F1 Score'))
pyo.iplot(go.Figure(data=traces, layout=layout)) | code |
2036047/cell_5 | [
"text_plain_output_1.png"
] | import pandas as pd
data_df = pd.read_csv('../input/mushrooms.csv')
data_df.head() | code |
32070789/cell_21 | [
"text_plain_output_2.png",
"text_plain_output_1.png",
"image_output_1.png"
] | from IPython import display
from keras.callbacks import ModelCheckpoint,ReduceLROnPlateau,EarlyStopping
from keras.layers import Conv2D,MaxPooling2D,GlobalMaxPooling2D
from keras.layers import Dense,LSTM,GlobalAveragePooling1D,GlobalAveragePooling2D
from keras.layers import Input,Activation,Flatten,Dropout,BatchNormalization
from keras.layers.advanced_activations import PReLU,LeakyReLU
from keras.metrics import top_k_categorical_accuracy,categorical_accuracy
from keras.models import Sequential,load_model,Model
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
import warnings
warnings.filterwarnings('ignore')
import numpy as np, pandas as pd, pylab as pl, h5py
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
from IPython import display
from keras.preprocessing.image import ImageDataGenerator
from keras.callbacks import ModelCheckpoint, ReduceLROnPlateau, EarlyStopping
from keras.metrics import top_k_categorical_accuracy, categorical_accuracy
from keras.models import Sequential, load_model, Model
from keras.layers import Dense, LSTM, GlobalAveragePooling1D, GlobalAveragePooling2D
from keras.layers.advanced_activations import PReLU, LeakyReLU
from keras.layers import Input, Activation, Flatten, Dropout, BatchNormalization
from keras.layers import Conv2D, MaxPooling2D, GlobalMaxPooling2D
fw = 'weights.best.letters.hdf5'
def history_plot(fit_history):
keys = list(fit_history.history.keys())[0:4]
def ohe(x):
return OneHotEncoder(categories='auto').fit(x.reshape(-1, 1)).transform(x.reshape(-1, 1)).toarray().astype('int64')
def tts(X, y):
x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)
n = int(len(x_test) / 2)
x_valid, y_valid = (x_test[:n], y_test[:n])
x_test, y_test = (x_test[n:], y_test[n:])
return (x_train, x_valid, x_test, y_train, y_valid, y_test)
f = h5py.File('../input/LetterColorImages_123.h5', 'r')
keys = list(f.keys())
keys
letters = u'абвгдеёжзийклмнопрстуфхцчшщъыьэюя'
backgrounds = np.array(f[keys[0]])
labels = np.array(f[keys[2]])
images = np.array(f[keys[1]]) / 255
gray_images = np.dot(images[..., :3], [0.299, 0.587, 0.114])
rn = np.random.randint(10000)
pl.xticks([])
pl.yticks([])
gray_images = gray_images.reshape(-1, 32, 32, 1)
cbackgrounds, clabels = (ohe(backgrounds), ohe(labels))
ctargets = np.concatenate((clabels, cbackgrounds), axis=1)
pd.DataFrame([clabels.shape, cbackgrounds.shape, ctargets.shape])
x_train1, x_valid1, x_test1, y_train1, y_valid1, y_test1 = tts(gray_images, clabels)
x_train2, x_valid2, x_test2, y_train2, y_valid2, y_test2 = tts(gray_images, ctargets)
y_train2_list = [y_train2[:, :33], y_train2[:, 33:]]
y_test2_list = [y_test2[:, :33], y_test2[:, 33:]]
y_valid2_list = [y_valid2[:, :33], y_valid2[:, 33:]]
def top_3_categorical_accuracy(y_true, y_pred):
return top_k_categorical_accuracy(y_true, y_pred, k=3)
def gray_model():
model = Sequential()
model.add(Conv2D(32, (5, 5), padding='same', input_shape=x_train1.shape[1:]))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(128, (5, 5)))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(GlobalMaxPooling2D())
model.add(Dense(1024))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(128))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(33))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=[categorical_accuracy, top_3_categorical_accuracy])
return model
gray_model = gray_model()
checkpointer = ModelCheckpoint(filepath=fw, verbose=2, save_best_only=True)
lr_reduction = ReduceLROnPlateau(monitor='val_loss', patience=10, verbose=2, factor=0.5)
estopping = EarlyStopping(monitor='val_loss', patience=20, verbose=2)
history = gray_model.fit(x_train1, y_train1, epochs=200, batch_size=64, verbose=2, validation_data=(x_valid1, y_valid1), callbacks=[checkpointer, lr_reduction, estopping])
def gray_multi_model():
model_input = Input(shape=(32, 32, 1))
x = BatchNormalization()(model_input)
x = Conv2D(32, (5, 5), padding='same')(model_input)
x = LeakyReLU(alpha=0.02)(x)
x = MaxPooling2D(pool_size=(2, 2))(x)
x = Dropout(0.25)(x)
x = Conv2D(256, (5, 5), padding='same')(x)
x = LeakyReLU(alpha=0.02)(x)
x = MaxPooling2D(pool_size=(2, 2))(x)
x = Dropout(0.25)(x)
x = GlobalMaxPooling2D()(x)
x = Dense(1024)(x)
x = LeakyReLU(alpha=0.02)(x)
x = Dropout(0.25)(x)
x = Dense(256)(x)
x = LeakyReLU(alpha=0.02)(x)
x = Dropout(0.025)(x)
y1 = Dense(33, activation='softmax')(x)
y2 = Dense(4, activation='softmax')(x)
model = Model(inputs=model_input, outputs=[y1, y2])
model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=[categorical_accuracy, top_3_categorical_accuracy])
return model
gray_multi_model = gray_multi_model()
checkpointer = ModelCheckpoint(filepath=fw, verbose=2, save_best_only=True)
lr_reduction = ReduceLROnPlateau(monitor='val_loss', patience=5, verbose=2, factor=0.75)
estopping = EarlyStopping(monitor='val_loss', patience=20, verbose=2)
history = gray_multi_model.fit(x_train2, y_train2_list, epochs=200, batch_size=128, verbose=2, validation_data=(x_valid2, y_valid2_list), callbacks=[checkpointer, lr_reduction, estopping])
gray_multi_model.evaluate(x_test2, y_test2_list) | code |
32070789/cell_13 | [
"text_plain_output_1.png"
] | from keras.callbacks import ModelCheckpoint,ReduceLROnPlateau,EarlyStopping
from keras.layers import Conv2D,MaxPooling2D,GlobalMaxPooling2D
from keras.layers import Dense,LSTM,GlobalAveragePooling1D,GlobalAveragePooling2D
from keras.layers import Input,Activation,Flatten,Dropout,BatchNormalization
from keras.layers.advanced_activations import PReLU,LeakyReLU
from keras.metrics import top_k_categorical_accuracy,categorical_accuracy
from keras.models import Sequential,load_model,Model
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
import warnings
warnings.filterwarnings('ignore')
import numpy as np, pandas as pd, pylab as pl, h5py
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
from IPython import display
from keras.preprocessing.image import ImageDataGenerator
from keras.callbacks import ModelCheckpoint, ReduceLROnPlateau, EarlyStopping
from keras.metrics import top_k_categorical_accuracy, categorical_accuracy
from keras.models import Sequential, load_model, Model
from keras.layers import Dense, LSTM, GlobalAveragePooling1D, GlobalAveragePooling2D
from keras.layers.advanced_activations import PReLU, LeakyReLU
from keras.layers import Input, Activation, Flatten, Dropout, BatchNormalization
from keras.layers import Conv2D, MaxPooling2D, GlobalMaxPooling2D
fw = 'weights.best.letters.hdf5'
def history_plot(fit_history):
keys = list(fit_history.history.keys())[0:4]
def ohe(x):
return OneHotEncoder(categories='auto').fit(x.reshape(-1, 1)).transform(x.reshape(-1, 1)).toarray().astype('int64')
def tts(X, y):
x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)
n = int(len(x_test) / 2)
x_valid, y_valid = (x_test[:n], y_test[:n])
x_test, y_test = (x_test[n:], y_test[n:])
return (x_train, x_valid, x_test, y_train, y_valid, y_test)
def top_3_categorical_accuracy(y_true, y_pred):
return top_k_categorical_accuracy(y_true, y_pred, k=3)
def gray_model():
model = Sequential()
model.add(Conv2D(32, (5, 5), padding='same', input_shape=x_train1.shape[1:]))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(128, (5, 5)))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(GlobalMaxPooling2D())
model.add(Dense(1024))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(128))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(33))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=[categorical_accuracy, top_3_categorical_accuracy])
return model
gray_model = gray_model()
checkpointer = ModelCheckpoint(filepath=fw, verbose=2, save_best_only=True)
lr_reduction = ReduceLROnPlateau(monitor='val_loss', patience=10, verbose=2, factor=0.5)
estopping = EarlyStopping(monitor='val_loss', patience=20, verbose=2)
history = gray_model.fit(x_train1, y_train1, epochs=200, batch_size=64, verbose=2, validation_data=(x_valid1, y_valid1), callbacks=[checkpointer, lr_reduction, estopping])
history_plot(history)
gray_model.load_weights(fw)
gray_model.evaluate(x_test1, y_test1) | code |
32070789/cell_20 | [
"text_plain_output_1.png"
] | from IPython import display
from keras.callbacks import ModelCheckpoint,ReduceLROnPlateau,EarlyStopping
from keras.layers import Conv2D,MaxPooling2D,GlobalMaxPooling2D
from keras.layers import Dense,LSTM,GlobalAveragePooling1D,GlobalAveragePooling2D
from keras.layers import Input,Activation,Flatten,Dropout,BatchNormalization
from keras.layers.advanced_activations import PReLU,LeakyReLU
from keras.metrics import top_k_categorical_accuracy,categorical_accuracy
from keras.models import Sequential,load_model,Model
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
import warnings
warnings.filterwarnings('ignore')
import numpy as np, pandas as pd, pylab as pl, h5py
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
from IPython import display
from keras.preprocessing.image import ImageDataGenerator
from keras.callbacks import ModelCheckpoint, ReduceLROnPlateau, EarlyStopping
from keras.metrics import top_k_categorical_accuracy, categorical_accuracy
from keras.models import Sequential, load_model, Model
from keras.layers import Dense, LSTM, GlobalAveragePooling1D, GlobalAveragePooling2D
from keras.layers.advanced_activations import PReLU, LeakyReLU
from keras.layers import Input, Activation, Flatten, Dropout, BatchNormalization
from keras.layers import Conv2D, MaxPooling2D, GlobalMaxPooling2D
fw = 'weights.best.letters.hdf5'
def history_plot(fit_history):
keys = list(fit_history.history.keys())[0:4]
def ohe(x):
return OneHotEncoder(categories='auto').fit(x.reshape(-1, 1)).transform(x.reshape(-1, 1)).toarray().astype('int64')
def tts(X, y):
x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)
n = int(len(x_test) / 2)
x_valid, y_valid = (x_test[:n], y_test[:n])
x_test, y_test = (x_test[n:], y_test[n:])
return (x_train, x_valid, x_test, y_train, y_valid, y_test)
f = h5py.File('../input/LetterColorImages_123.h5', 'r')
keys = list(f.keys())
keys
letters = u'абвгдеёжзийклмнопрстуфхцчшщъыьэюя'
backgrounds = np.array(f[keys[0]])
labels = np.array(f[keys[2]])
images = np.array(f[keys[1]]) / 255
gray_images = np.dot(images[..., :3], [0.299, 0.587, 0.114])
rn = np.random.randint(10000)
pl.xticks([])
pl.yticks([])
gray_images = gray_images.reshape(-1, 32, 32, 1)
cbackgrounds, clabels = (ohe(backgrounds), ohe(labels))
ctargets = np.concatenate((clabels, cbackgrounds), axis=1)
pd.DataFrame([clabels.shape, cbackgrounds.shape, ctargets.shape])
x_train1, x_valid1, x_test1, y_train1, y_valid1, y_test1 = tts(gray_images, clabels)
x_train2, x_valid2, x_test2, y_train2, y_valid2, y_test2 = tts(gray_images, ctargets)
y_train2_list = [y_train2[:, :33], y_train2[:, 33:]]
y_test2_list = [y_test2[:, :33], y_test2[:, 33:]]
y_valid2_list = [y_valid2[:, :33], y_valid2[:, 33:]]
def top_3_categorical_accuracy(y_true, y_pred):
return top_k_categorical_accuracy(y_true, y_pred, k=3)
def gray_model():
model = Sequential()
model.add(Conv2D(32, (5, 5), padding='same', input_shape=x_train1.shape[1:]))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(128, (5, 5)))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(GlobalMaxPooling2D())
model.add(Dense(1024))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(128))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(33))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=[categorical_accuracy, top_3_categorical_accuracy])
return model
gray_model = gray_model()
checkpointer = ModelCheckpoint(filepath=fw, verbose=2, save_best_only=True)
lr_reduction = ReduceLROnPlateau(monitor='val_loss', patience=10, verbose=2, factor=0.5)
estopping = EarlyStopping(monitor='val_loss', patience=20, verbose=2)
history = gray_model.fit(x_train1, y_train1, epochs=200, batch_size=64, verbose=2, validation_data=(x_valid1, y_valid1), callbacks=[checkpointer, lr_reduction, estopping])
def gray_multi_model():
model_input = Input(shape=(32, 32, 1))
x = BatchNormalization()(model_input)
x = Conv2D(32, (5, 5), padding='same')(model_input)
x = LeakyReLU(alpha=0.02)(x)
x = MaxPooling2D(pool_size=(2, 2))(x)
x = Dropout(0.25)(x)
x = Conv2D(256, (5, 5), padding='same')(x)
x = LeakyReLU(alpha=0.02)(x)
x = MaxPooling2D(pool_size=(2, 2))(x)
x = Dropout(0.25)(x)
x = GlobalMaxPooling2D()(x)
x = Dense(1024)(x)
x = LeakyReLU(alpha=0.02)(x)
x = Dropout(0.25)(x)
x = Dense(256)(x)
x = LeakyReLU(alpha=0.02)(x)
x = Dropout(0.025)(x)
y1 = Dense(33, activation='softmax')(x)
y2 = Dense(4, activation='softmax')(x)
model = Model(inputs=model_input, outputs=[y1, y2])
model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=[categorical_accuracy, top_3_categorical_accuracy])
return model
gray_multi_model = gray_multi_model()
checkpointer = ModelCheckpoint(filepath=fw, verbose=2, save_best_only=True)
lr_reduction = ReduceLROnPlateau(monitor='val_loss', patience=5, verbose=2, factor=0.75)
estopping = EarlyStopping(monitor='val_loss', patience=20, verbose=2)
history = gray_multi_model.fit(x_train2, y_train2_list, epochs=200, batch_size=128, verbose=2, validation_data=(x_valid2, y_valid2_list), callbacks=[checkpointer, lr_reduction, estopping]) | code |
32070789/cell_2 | [
"text_plain_output_2.png",
"text_plain_output_1.png",
"image_output_1.png"
] | import warnings
warnings.filterwarnings('ignore')
import numpy as np, pandas as pd, pylab as pl, h5py
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
from IPython import display
from keras.preprocessing.image import ImageDataGenerator
from keras.callbacks import ModelCheckpoint, ReduceLROnPlateau, EarlyStopping
from keras.metrics import top_k_categorical_accuracy, categorical_accuracy
from keras.models import Sequential, load_model, Model
from keras.layers import Dense, LSTM, GlobalAveragePooling1D, GlobalAveragePooling2D
from keras.layers.advanced_activations import PReLU, LeakyReLU
from keras.layers import Input, Activation, Flatten, Dropout, BatchNormalization
from keras.layers import Conv2D, MaxPooling2D, GlobalMaxPooling2D
fw = 'weights.best.letters.hdf5' | code |
32070789/cell_7 | [
"text_plain_output_1.png"
] | from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
def history_plot(fit_history):
keys = list(fit_history.history.keys())[0:4]
def ohe(x):
return OneHotEncoder(categories='auto').fit(x.reshape(-1, 1)).transform(x.reshape(-1, 1)).toarray().astype('int64')
def tts(X, y):
x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)
n = int(len(x_test) / 2)
x_valid, y_valid = (x_test[:n], y_test[:n])
x_test, y_test = (x_test[n:], y_test[n:])
return (x_train, x_valid, x_test, y_train, y_valid, y_test)
f = h5py.File('../input/LetterColorImages_123.h5', 'r')
keys = list(f.keys())
keys
letters = u'абвгдеёжзийклмнопрстуфхцчшщъыьэюя'
backgrounds = np.array(f[keys[0]])
labels = np.array(f[keys[2]])
images = np.array(f[keys[1]]) / 255
gray_images = np.dot(images[..., :3], [0.299, 0.587, 0.114])
rn = np.random.randint(10000)
pl.figure(figsize=(2, 3))
pl.title('Label: %s \n' % letters[labels[rn] - 1] + 'Background: %s' % backgrounds[rn], fontsize=18)
pl.imshow(gray_images[rn], cmap=pl.cm.bone)
pl.xticks([])
pl.yticks([])
pl.show()
gray_images = gray_images.reshape(-1, 32, 32, 1) | code |
32070789/cell_8 | [
"text_plain_output_2.png",
"text_plain_output_1.png"
] | from IPython import display
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
def history_plot(fit_history):
keys = list(fit_history.history.keys())[0:4]
def ohe(x):
return OneHotEncoder(categories='auto').fit(x.reshape(-1, 1)).transform(x.reshape(-1, 1)).toarray().astype('int64')
def tts(X, y):
x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)
n = int(len(x_test) / 2)
x_valid, y_valid = (x_test[:n], y_test[:n])
x_test, y_test = (x_test[n:], y_test[n:])
return (x_train, x_valid, x_test, y_train, y_valid, y_test)
f = h5py.File('../input/LetterColorImages_123.h5', 'r')
keys = list(f.keys())
keys
letters = u'абвгдеёжзийклмнопрстуфхцчшщъыьэюя'
backgrounds = np.array(f[keys[0]])
labels = np.array(f[keys[2]])
images = np.array(f[keys[1]]) / 255
gray_images = np.dot(images[..., :3], [0.299, 0.587, 0.114])
rn = np.random.randint(10000)
pl.xticks([])
pl.yticks([])
gray_images = gray_images.reshape(-1, 32, 32, 1)
cbackgrounds, clabels = (ohe(backgrounds), ohe(labels))
ctargets = np.concatenate((clabels, cbackgrounds), axis=1)
display.display(pd.DataFrame([labels[97:103], clabels[97:103]]).T)
pd.DataFrame([clabels.shape, cbackgrounds.shape, ctargets.shape]) | code |
32070789/cell_15 | [
"text_html_output_2.png",
"text_html_output_1.png"
] | from keras.callbacks import ModelCheckpoint,ReduceLROnPlateau,EarlyStopping
from keras.layers import Conv2D,MaxPooling2D,GlobalMaxPooling2D
from keras.layers import Dense,LSTM,GlobalAveragePooling1D,GlobalAveragePooling2D
from keras.layers import Input,Activation,Flatten,Dropout,BatchNormalization
from keras.layers.advanced_activations import PReLU,LeakyReLU
from keras.metrics import top_k_categorical_accuracy,categorical_accuracy
from keras.models import Sequential,load_model,Model
from keras.preprocessing.image import ImageDataGenerator
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
import warnings
warnings.filterwarnings('ignore')
import numpy as np, pandas as pd, pylab as pl, h5py
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
from IPython import display
from keras.preprocessing.image import ImageDataGenerator
from keras.callbacks import ModelCheckpoint, ReduceLROnPlateau, EarlyStopping
from keras.metrics import top_k_categorical_accuracy, categorical_accuracy
from keras.models import Sequential, load_model, Model
from keras.layers import Dense, LSTM, GlobalAveragePooling1D, GlobalAveragePooling2D
from keras.layers.advanced_activations import PReLU, LeakyReLU
from keras.layers import Input, Activation, Flatten, Dropout, BatchNormalization
from keras.layers import Conv2D, MaxPooling2D, GlobalMaxPooling2D
fw = 'weights.best.letters.hdf5'
def history_plot(fit_history):
keys = list(fit_history.history.keys())[0:4]
def ohe(x):
return OneHotEncoder(categories='auto').fit(x.reshape(-1, 1)).transform(x.reshape(-1, 1)).toarray().astype('int64')
def tts(X, y):
x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)
n = int(len(x_test) / 2)
x_valid, y_valid = (x_test[:n], y_test[:n])
x_test, y_test = (x_test[n:], y_test[n:])
return (x_train, x_valid, x_test, y_train, y_valid, y_test)
def top_3_categorical_accuracy(y_true, y_pred):
return top_k_categorical_accuracy(y_true, y_pred, k=3)
def gray_model():
model = Sequential()
model.add(Conv2D(32, (5, 5), padding='same', input_shape=x_train1.shape[1:]))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(128, (5, 5)))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(GlobalMaxPooling2D())
model.add(Dense(1024))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(128))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(33))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=[categorical_accuracy, top_3_categorical_accuracy])
return model
gray_model = gray_model()
checkpointer = ModelCheckpoint(filepath=fw, verbose=2, save_best_only=True)
lr_reduction = ReduceLROnPlateau(monitor='val_loss', patience=10, verbose=2, factor=0.5)
estopping = EarlyStopping(monitor='val_loss', patience=20, verbose=2)
history = gray_model.fit(x_train1, y_train1, epochs=200, batch_size=64, verbose=2, validation_data=(x_valid1, y_valid1), callbacks=[checkpointer, lr_reduction, estopping])
gray_model.load_weights(fw)
gray_model.evaluate(x_test1, y_test1)
steps, epochs = (1000, 10)
igen = ImageDataGenerator(zoom_range=0.3, shear_range=0.3, rotation_range=30)
generator = gray_model.fit_generator(igen.flow(x_train1, y_train1, batch_size=64), steps_per_epoch=steps, epochs=epochs, verbose=2, validation_data=(x_valid1, y_valid1), callbacks=[checkpointer, lr_reduction])
history_plot(generator)
gray_model.load_weights(fw)
gray_model.evaluate(x_test1, y_test1) | code |
32070789/cell_17 | [
"text_plain_output_1.png"
] | from IPython import display
from keras.callbacks import ModelCheckpoint,ReduceLROnPlateau,EarlyStopping
from keras.layers import Conv2D,MaxPooling2D,GlobalMaxPooling2D
from keras.layers import Dense,LSTM,GlobalAveragePooling1D,GlobalAveragePooling2D
from keras.layers import Input,Activation,Flatten,Dropout,BatchNormalization
from keras.layers.advanced_activations import PReLU,LeakyReLU
from keras.metrics import top_k_categorical_accuracy,categorical_accuracy
from keras.models import Sequential,load_model,Model
from keras.preprocessing.image import ImageDataGenerator
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
import warnings
warnings.filterwarnings('ignore')
import numpy as np, pandas as pd, pylab as pl, h5py
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
from IPython import display
from keras.preprocessing.image import ImageDataGenerator
from keras.callbacks import ModelCheckpoint, ReduceLROnPlateau, EarlyStopping
from keras.metrics import top_k_categorical_accuracy, categorical_accuracy
from keras.models import Sequential, load_model, Model
from keras.layers import Dense, LSTM, GlobalAveragePooling1D, GlobalAveragePooling2D
from keras.layers.advanced_activations import PReLU, LeakyReLU
from keras.layers import Input, Activation, Flatten, Dropout, BatchNormalization
from keras.layers import Conv2D, MaxPooling2D, GlobalMaxPooling2D
fw = 'weights.best.letters.hdf5'
def history_plot(fit_history):
keys = list(fit_history.history.keys())[0:4]
def ohe(x):
return OneHotEncoder(categories='auto').fit(x.reshape(-1, 1)).transform(x.reshape(-1, 1)).toarray().astype('int64')
def tts(X, y):
x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)
n = int(len(x_test) / 2)
x_valid, y_valid = (x_test[:n], y_test[:n])
x_test, y_test = (x_test[n:], y_test[n:])
return (x_train, x_valid, x_test, y_train, y_valid, y_test)
f = h5py.File('../input/LetterColorImages_123.h5', 'r')
keys = list(f.keys())
keys
letters = u'абвгдеёжзийклмнопрстуфхцчшщъыьэюя'
backgrounds = np.array(f[keys[0]])
labels = np.array(f[keys[2]])
images = np.array(f[keys[1]]) / 255
gray_images = np.dot(images[..., :3], [0.299, 0.587, 0.114])
rn = np.random.randint(10000)
pl.xticks([])
pl.yticks([])
gray_images = gray_images.reshape(-1, 32, 32, 1)
cbackgrounds, clabels = (ohe(backgrounds), ohe(labels))
ctargets = np.concatenate((clabels, cbackgrounds), axis=1)
pd.DataFrame([clabels.shape, cbackgrounds.shape, ctargets.shape])
def top_3_categorical_accuracy(y_true, y_pred):
return top_k_categorical_accuracy(y_true, y_pred, k=3)
def gray_model():
model = Sequential()
model.add(Conv2D(32, (5, 5), padding='same', input_shape=x_train1.shape[1:]))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(128, (5, 5)))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(GlobalMaxPooling2D())
model.add(Dense(1024))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(128))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(33))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=[categorical_accuracy, top_3_categorical_accuracy])
return model
gray_model = gray_model()
checkpointer = ModelCheckpoint(filepath=fw, verbose=2, save_best_only=True)
lr_reduction = ReduceLROnPlateau(monitor='val_loss', patience=10, verbose=2, factor=0.5)
estopping = EarlyStopping(monitor='val_loss', patience=20, verbose=2)
history = gray_model.fit(x_train1, y_train1, epochs=200, batch_size=64, verbose=2, validation_data=(x_valid1, y_valid1), callbacks=[checkpointer, lr_reduction, estopping])
gray_model.load_weights(fw)
gray_model.evaluate(x_test1, y_test1)
steps, epochs = (1000, 10)
igen = ImageDataGenerator(zoom_range=0.3, shear_range=0.3, rotation_range=30)
generator = gray_model.fit_generator(igen.flow(x_train1, y_train1, batch_size=64), steps_per_epoch=steps, epochs=epochs, verbose=2, validation_data=(x_valid1, y_valid1), callbacks=[checkpointer, lr_reduction])
gray_model.load_weights(fw)
gray_model.evaluate(x_test1, y_test1)
py_test1 = gray_model.predict_classes(x_test1)
fig = pl.figure(figsize=(12, 12))
for i, idx in enumerate(np.random.choice(x_test1.shape[0], size=16, replace=False)):
ax = fig.add_subplot(4, 4, i + 1, xticks=[], yticks=[])
ax.imshow(np.squeeze(x_test1[idx]), cmap=pl.cm.bone)
pred_idx = py_test1[idx]
true_idx = np.argmax(y_test1[idx])
ax.set_title('{} ({})'.format(letters[pred_idx], letters[true_idx]), color='darkblue' if pred_idx == true_idx else 'darkred') | code |
32070789/cell_14 | [
"image_output_1.png"
] | from keras.callbacks import ModelCheckpoint,ReduceLROnPlateau,EarlyStopping
from keras.layers import Conv2D,MaxPooling2D,GlobalMaxPooling2D
from keras.layers import Dense,LSTM,GlobalAveragePooling1D,GlobalAveragePooling2D
from keras.layers import Input,Activation,Flatten,Dropout,BatchNormalization
from keras.layers.advanced_activations import PReLU,LeakyReLU
from keras.metrics import top_k_categorical_accuracy,categorical_accuracy
from keras.models import Sequential,load_model,Model
from keras.preprocessing.image import ImageDataGenerator
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
import warnings
warnings.filterwarnings('ignore')
import numpy as np, pandas as pd, pylab as pl, h5py
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
from IPython import display
from keras.preprocessing.image import ImageDataGenerator
from keras.callbacks import ModelCheckpoint, ReduceLROnPlateau, EarlyStopping
from keras.metrics import top_k_categorical_accuracy, categorical_accuracy
from keras.models import Sequential, load_model, Model
from keras.layers import Dense, LSTM, GlobalAveragePooling1D, GlobalAveragePooling2D
from keras.layers.advanced_activations import PReLU, LeakyReLU
from keras.layers import Input, Activation, Flatten, Dropout, BatchNormalization
from keras.layers import Conv2D, MaxPooling2D, GlobalMaxPooling2D
fw = 'weights.best.letters.hdf5'
def history_plot(fit_history):
keys = list(fit_history.history.keys())[0:4]
def ohe(x):
return OneHotEncoder(categories='auto').fit(x.reshape(-1, 1)).transform(x.reshape(-1, 1)).toarray().astype('int64')
def tts(X, y):
x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)
n = int(len(x_test) / 2)
x_valid, y_valid = (x_test[:n], y_test[:n])
x_test, y_test = (x_test[n:], y_test[n:])
return (x_train, x_valid, x_test, y_train, y_valid, y_test)
def top_3_categorical_accuracy(y_true, y_pred):
return top_k_categorical_accuracy(y_true, y_pred, k=3)
def gray_model():
model = Sequential()
model.add(Conv2D(32, (5, 5), padding='same', input_shape=x_train1.shape[1:]))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(128, (5, 5)))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(GlobalMaxPooling2D())
model.add(Dense(1024))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(128))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(33))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=[categorical_accuracy, top_3_categorical_accuracy])
return model
gray_model = gray_model()
checkpointer = ModelCheckpoint(filepath=fw, verbose=2, save_best_only=True)
lr_reduction = ReduceLROnPlateau(monitor='val_loss', patience=10, verbose=2, factor=0.5)
estopping = EarlyStopping(monitor='val_loss', patience=20, verbose=2)
history = gray_model.fit(x_train1, y_train1, epochs=200, batch_size=64, verbose=2, validation_data=(x_valid1, y_valid1), callbacks=[checkpointer, lr_reduction, estopping])
gray_model.load_weights(fw)
gray_model.evaluate(x_test1, y_test1)
steps, epochs = (1000, 10)
igen = ImageDataGenerator(zoom_range=0.3, shear_range=0.3, rotation_range=30)
generator = gray_model.fit_generator(igen.flow(x_train1, y_train1, batch_size=64), steps_per_epoch=steps, epochs=epochs, verbose=2, validation_data=(x_valid1, y_valid1), callbacks=[checkpointer, lr_reduction]) | code |
32070789/cell_12 | [
"application_vnd.jupyter.stderr_output_1.png"
] | from keras.callbacks import ModelCheckpoint,ReduceLROnPlateau,EarlyStopping
from keras.layers import Conv2D,MaxPooling2D,GlobalMaxPooling2D
from keras.layers import Dense,LSTM,GlobalAveragePooling1D,GlobalAveragePooling2D
from keras.layers import Input,Activation,Flatten,Dropout,BatchNormalization
from keras.layers.advanced_activations import PReLU,LeakyReLU
from keras.metrics import top_k_categorical_accuracy,categorical_accuracy
from keras.models import Sequential,load_model,Model
import warnings
warnings.filterwarnings('ignore')
import numpy as np, pandas as pd, pylab as pl, h5py
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
from IPython import display
from keras.preprocessing.image import ImageDataGenerator
from keras.callbacks import ModelCheckpoint, ReduceLROnPlateau, EarlyStopping
from keras.metrics import top_k_categorical_accuracy, categorical_accuracy
from keras.models import Sequential, load_model, Model
from keras.layers import Dense, LSTM, GlobalAveragePooling1D, GlobalAveragePooling2D
from keras.layers.advanced_activations import PReLU, LeakyReLU
from keras.layers import Input, Activation, Flatten, Dropout, BatchNormalization
from keras.layers import Conv2D, MaxPooling2D, GlobalMaxPooling2D
fw = 'weights.best.letters.hdf5'
def top_3_categorical_accuracy(y_true, y_pred):
return top_k_categorical_accuracy(y_true, y_pred, k=3)
def gray_model():
model = Sequential()
model.add(Conv2D(32, (5, 5), padding='same', input_shape=x_train1.shape[1:]))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(128, (5, 5)))
model.add(LeakyReLU(alpha=0.02))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(GlobalMaxPooling2D())
model.add(Dense(1024))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(128))
model.add(LeakyReLU(alpha=0.02))
model.add(Dropout(0.25))
model.add(Dense(33))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=[categorical_accuracy, top_3_categorical_accuracy])
return model
gray_model = gray_model()
checkpointer = ModelCheckpoint(filepath=fw, verbose=2, save_best_only=True)
lr_reduction = ReduceLROnPlateau(monitor='val_loss', patience=10, verbose=2, factor=0.5)
estopping = EarlyStopping(monitor='val_loss', patience=20, verbose=2)
history = gray_model.fit(x_train1, y_train1, epochs=200, batch_size=64, verbose=2, validation_data=(x_valid1, y_valid1), callbacks=[checkpointer, lr_reduction, estopping]) | code |
32070789/cell_5 | [
"image_output_1.png"
] | from sklearn.model_selection import train_test_split
from sklearn.preprocessing import OneHotEncoder
def history_plot(fit_history):
keys = list(fit_history.history.keys())[0:4]
def ohe(x):
return OneHotEncoder(categories='auto').fit(x.reshape(-1, 1)).transform(x.reshape(-1, 1)).toarray().astype('int64')
def tts(X, y):
x_train, x_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=1)
n = int(len(x_test) / 2)
x_valid, y_valid = (x_test[:n], y_test[:n])
x_test, y_test = (x_test[n:], y_test[n:])
return (x_train, x_valid, x_test, y_train, y_valid, y_test)
f = h5py.File('../input/LetterColorImages_123.h5', 'r')
keys = list(f.keys())
keys | code |
16130893/cell_4 | [
"text_plain_output_1.png",
"image_output_1.png"
] | import cv2
import matplotlib.pyplot as plt
import numpy as np
import os
import numpy as np
import pandas as pd
import cv2
from IPython.display import Image
import matplotlib.pyplot as plt
import os
import tensorflow as tf
from tensorflow.keras import layers
from sklearn.model_selection import train_test_split
onlyfiles = os.listdir('../input/utkface_aligned_cropped/UTKFace')
y = np.array([[[i.split('_')[0]], [i.split('_')[1]]] for i in onlyfiles])
X_data = []
for file in onlyfiles:
face = cv2.imread('../input/utkface_aligned_cropped/UTKFace/' + file, cv2.IMREAD_COLOR)
face = cv2.cvtColor(face, cv2.COLOR_BGR2RGB)
face = cv2.resize(face, (32, 32))
X_data.append(face)
X_data = np.array(X_data)
X_data.shape
X = np.squeeze(X_data)
plt.imshow(X[1], interpolation='bicubic')
(plt.xticks([]), plt.yticks([]))
plt.show()
print(y[1]) | code |
16130893/cell_6 | [
"image_output_1.png"
] | from IPython.display import Image
from tensorflow.keras import layers
import tensorflow as tf
def gen_model():
inputs = tf.keras.layers.Input(shape=(32, 32, 3))
x = inputs
x = layers.Conv2D(32, 3, activation='relu')(x)
x = layers.Conv2D(32, 3, activation='relu')(x)
x = layers.MaxPool2D(2)(x)
x = layers.Dropout(0.3)(x)
x = layers.Conv2D(64, 3, activation='relu')(x)
x = layers.Conv2D(64, 3, activation='relu')(x)
x = layers.MaxPool2D(2)(x)
x = layers.Dropout(0.3)(x)
x = layers.Conv2D(84, 3, activation='relu')(x)
x = layers.Dropout(0.3)(x)
x = layers.Flatten()(x)
x1 = layers.Dense(64, activation='relu')(x)
x2 = layers.Dense(64, activation='relu')(x)
x1 = layers.Dense(1, activation='sigmoid', name='sex_out')(x1)
x2 = layers.Dense(1, activation='relu', name='age_out')(x2)
model = tf.keras.models.Model(inputs=inputs, outputs=[x1, x2])
model.compile(optimizer='Adam', loss=['binary_crossentropy', 'mae'])
tf.keras.utils.plot_model(model, 'model.png', show_shapes=True)
return model
model = gen_model()
Image('model.png') | code |
16130893/cell_2 | [
"text_plain_output_1.png"
] | import numpy as np
import os
import numpy as np
import pandas as pd
import cv2
from IPython.display import Image
import matplotlib.pyplot as plt
import os
import tensorflow as tf
from tensorflow.keras import layers
from sklearn.model_selection import train_test_split
onlyfiles = os.listdir('../input/utkface_aligned_cropped/UTKFace')
y = np.array([[[i.split('_')[0]], [i.split('_')[1]]] for i in onlyfiles])
print(y.shape)
print(y[0]) | code |
16130893/cell_1 | [
"text_plain_output_1.png"
] | import os
import numpy as np
import pandas as pd
import cv2
from IPython.display import Image
import matplotlib.pyplot as plt
import os
import tensorflow as tf
from tensorflow.keras import layers
from sklearn.model_selection import train_test_split
print(os.listdir('../input/utkface_aligned_cropped/')) | code |
16130893/cell_7 | [
"text_plain_output_2.png",
"text_plain_output_1.png"
] | from IPython.display import Image
from sklearn.model_selection import train_test_split
from tensorflow.keras import layers
import cv2
import matplotlib.pyplot as plt
import numpy as np
import os
import tensorflow as tf
import numpy as np
import pandas as pd
import cv2
from IPython.display import Image
import matplotlib.pyplot as plt
import os
import tensorflow as tf
from tensorflow.keras import layers
from sklearn.model_selection import train_test_split
onlyfiles = os.listdir('../input/utkface_aligned_cropped/UTKFace')
y = np.array([[[i.split('_')[0]], [i.split('_')[1]]] for i in onlyfiles])
X_data = []
for file in onlyfiles:
face = cv2.imread('../input/utkface_aligned_cropped/UTKFace/' + file, cv2.IMREAD_COLOR)
face = cv2.cvtColor(face, cv2.COLOR_BGR2RGB)
face = cv2.resize(face, (32, 32))
X_data.append(face)
X_data = np.array(X_data)
X_data.shape
X = np.squeeze(X_data)
(plt.xticks([]), plt.yticks([]))
X_train, X_valid, y_train, y_valid = train_test_split(X, y, test_size=0.33)
y_train = [y_train[:, 1], y_train[:, 0]]
y_valid = [y_valid[:, 1], y_valid[:, 0]]
def gen_model():
inputs = tf.keras.layers.Input(shape=(32, 32, 3))
x = inputs
x = layers.Conv2D(32, 3, activation='relu')(x)
x = layers.Conv2D(32, 3, activation='relu')(x)
x = layers.MaxPool2D(2)(x)
x = layers.Dropout(0.3)(x)
x = layers.Conv2D(64, 3, activation='relu')(x)
x = layers.Conv2D(64, 3, activation='relu')(x)
x = layers.MaxPool2D(2)(x)
x = layers.Dropout(0.3)(x)
x = layers.Conv2D(84, 3, activation='relu')(x)
x = layers.Dropout(0.3)(x)
x = layers.Flatten()(x)
x1 = layers.Dense(64, activation='relu')(x)
x2 = layers.Dense(64, activation='relu')(x)
x1 = layers.Dense(1, activation='sigmoid', name='sex_out')(x1)
x2 = layers.Dense(1, activation='relu', name='age_out')(x2)
model = tf.keras.models.Model(inputs=inputs, outputs=[x1, x2])
model.compile(optimizer='Adam', loss=['binary_crossentropy', 'mae'])
return model
model = gen_model()
Image('model.png')
model.summary()
model.fit(X_train, y_train, epochs=200, batch_size=120, validation_data=(X_valid, y_valid)) | code |
16130893/cell_8 | [
"text_plain_output_1.png"
] | from IPython.display import Image
from sklearn.model_selection import train_test_split
from tensorflow.keras import layers
import cv2
import matplotlib.pyplot as plt
import numpy as np
import os
import tensorflow as tf
import numpy as np
import pandas as pd
import cv2
from IPython.display import Image
import matplotlib.pyplot as plt
import os
import tensorflow as tf
from tensorflow.keras import layers
from sklearn.model_selection import train_test_split
onlyfiles = os.listdir('../input/utkface_aligned_cropped/UTKFace')
y = np.array([[[i.split('_')[0]], [i.split('_')[1]]] for i in onlyfiles])
X_data = []
for file in onlyfiles:
face = cv2.imread('../input/utkface_aligned_cropped/UTKFace/' + file, cv2.IMREAD_COLOR)
face = cv2.cvtColor(face, cv2.COLOR_BGR2RGB)
face = cv2.resize(face, (32, 32))
X_data.append(face)
X_data = np.array(X_data)
X_data.shape
X = np.squeeze(X_data)
(plt.xticks([]), plt.yticks([]))
X_train, X_valid, y_train, y_valid = train_test_split(X, y, test_size=0.33)
y_train = [y_train[:, 1], y_train[:, 0]]
y_valid = [y_valid[:, 1], y_valid[:, 0]]
def gen_model():
inputs = tf.keras.layers.Input(shape=(32, 32, 3))
x = inputs
x = layers.Conv2D(32, 3, activation='relu')(x)
x = layers.Conv2D(32, 3, activation='relu')(x)
x = layers.MaxPool2D(2)(x)
x = layers.Dropout(0.3)(x)
x = layers.Conv2D(64, 3, activation='relu')(x)
x = layers.Conv2D(64, 3, activation='relu')(x)
x = layers.MaxPool2D(2)(x)
x = layers.Dropout(0.3)(x)
x = layers.Conv2D(84, 3, activation='relu')(x)
x = layers.Dropout(0.3)(x)
x = layers.Flatten()(x)
x1 = layers.Dense(64, activation='relu')(x)
x2 = layers.Dense(64, activation='relu')(x)
x1 = layers.Dense(1, activation='sigmoid', name='sex_out')(x1)
x2 = layers.Dense(1, activation='relu', name='age_out')(x2)
model = tf.keras.models.Model(inputs=inputs, outputs=[x1, x2])
model.compile(optimizer='Adam', loss=['binary_crossentropy', 'mae'])
return model
model = gen_model()
Image('model.png')
model.summary()
model.fit(X_train, y_train, epochs=200, batch_size=120, validation_data=(X_valid, y_valid))
print(y_valid[0][24], y_valid[1][24])
print(model.predict([[X_valid[24]]])) | code |
16130893/cell_3 | [
"text_plain_output_1.png"
] | import cv2
import numpy as np
import os
import numpy as np
import pandas as pd
import cv2
from IPython.display import Image
import matplotlib.pyplot as plt
import os
import tensorflow as tf
from tensorflow.keras import layers
from sklearn.model_selection import train_test_split
onlyfiles = os.listdir('../input/utkface_aligned_cropped/UTKFace')
y = np.array([[[i.split('_')[0]], [i.split('_')[1]]] for i in onlyfiles])
X_data = []
for file in onlyfiles:
face = cv2.imread('../input/utkface_aligned_cropped/UTKFace/' + file, cv2.IMREAD_COLOR)
face = cv2.cvtColor(face, cv2.COLOR_BGR2RGB)
face = cv2.resize(face, (32, 32))
X_data.append(face)
X_data = np.array(X_data)
X_data.shape | code |
90130201/cell_21 | [
"text_plain_output_1.png"
] | import pandas as pd
stroke_df = pd.read_csv('../input/stroke-prediction-dataset/healthcare-dataset-stroke-data.csv')
stroke_df.shape
na_count = []
for i in range(0, len(stroke_df.columns)):
na_count.append(stroke_df[stroke_df.columns[i]].isna().sum())
na_df = pd.DataFrame(zip(stroke_df.columns, na_count))
na_df.columns = ['Variable', 'Counts (NA)']
na_df
stroke_df_copy = stroke_df.copy()
pd.crosstab(stroke_df.loc[stroke_df['bmi'].isna()].gender, stroke_df.loc[stroke_df['bmi'].isna()].stroke, margins=True)
pd.crosstab(stroke_df.loc[stroke_df['bmi'].isna()].smoking_status, stroke_df.loc[stroke_df['bmi'].isna()].stroke, margins=True) | code |
90130201/cell_13 | [
"text_html_output_1.png"
] | 201 / 5110 | code |
90130201/cell_9 | [
"text_html_output_1.png"
] | import pandas as pd
stroke_df = pd.read_csv('../input/stroke-prediction-dataset/healthcare-dataset-stroke-data.csv')
stroke_df.shape | code |
90130201/cell_20 | [
"text_plain_output_1.png"
] | import pandas as pd
stroke_df = pd.read_csv('../input/stroke-prediction-dataset/healthcare-dataset-stroke-data.csv')
stroke_df.shape
na_count = []
for i in range(0, len(stroke_df.columns)):
na_count.append(stroke_df[stroke_df.columns[i]].isna().sum())
na_df = pd.DataFrame(zip(stroke_df.columns, na_count))
na_df.columns = ['Variable', 'Counts (NA)']
na_df
stroke_df_copy = stroke_df.copy()
pd.crosstab(stroke_df.loc[stroke_df['bmi'].isna()].gender, stroke_df.loc[stroke_df['bmi'].isna()].stroke, margins=True) | code |
90130201/cell_11 | [
"text_html_output_1.png"
] | import pandas as pd
stroke_df = pd.read_csv('../input/stroke-prediction-dataset/healthcare-dataset-stroke-data.csv')
stroke_df.shape
stroke_df.describe() | code |
90130201/cell_18 | [
"text_html_output_1.png"
] | import pandas as pd
stroke_df = pd.read_csv('../input/stroke-prediction-dataset/healthcare-dataset-stroke-data.csv')
stroke_df.shape
na_count = []
for i in range(0, len(stroke_df.columns)):
na_count.append(stroke_df[stroke_df.columns[i]].isna().sum())
na_df = pd.DataFrame(zip(stroke_df.columns, na_count))
na_df.columns = ['Variable', 'Counts (NA)']
na_df
for i in ['gender', 'hypertension', 'heart_disease', 'ever_married', 'work_type', 'Residence_type', 'smoking_status', 'stroke']:
print(stroke_df.loc[stroke_df['bmi'].isna()][i].value_counts(), '\n') | code |
90130201/cell_8 | [
"text_html_output_1.png"
] | import pandas as pd
stroke_df = pd.read_csv('../input/stroke-prediction-dataset/healthcare-dataset-stroke-data.csv')
stroke_df.head(5) | code |
90130201/cell_15 | [
"text_html_output_1.png"
] | import pandas as pd
stroke_df = pd.read_csv('../input/stroke-prediction-dataset/healthcare-dataset-stroke-data.csv')
stroke_df.shape
na_count = []
for i in range(0, len(stroke_df.columns)):
na_count.append(stroke_df[stroke_df.columns[i]].isna().sum())
na_df = pd.DataFrame(zip(stroke_df.columns, na_count))
na_df.columns = ['Variable', 'Counts (NA)']
na_df
stroke_df.loc[stroke_df['bmi'].isna()].describe() | code |
90130201/cell_16 | [
"text_plain_output_1.png"
] | import pandas as pd
stroke_df = pd.read_csv('../input/stroke-prediction-dataset/healthcare-dataset-stroke-data.csv')
stroke_df.shape
na_count = []
for i in range(0, len(stroke_df.columns)):
na_count.append(stroke_df[stroke_df.columns[i]].isna().sum())
na_df = pd.DataFrame(zip(stroke_df.columns, na_count))
na_df.columns = ['Variable', 'Counts (NA)']
na_df
for i in ['gender', 'hypertension', 'heart_disease', 'ever_married', 'work_type', 'Residence_type', 'smoking_status', 'stroke']:
print(stroke_df[i].value_counts(), '\n') | code |
90130201/cell_22 | [
"text_html_output_1.png"
] | import pandas as pd
stroke_df = pd.read_csv('../input/stroke-prediction-dataset/healthcare-dataset-stroke-data.csv')
stroke_df.shape
na_count = []
for i in range(0, len(stroke_df.columns)):
na_count.append(stroke_df[stroke_df.columns[i]].isna().sum())
na_df = pd.DataFrame(zip(stroke_df.columns, na_count))
na_df.columns = ['Variable', 'Counts (NA)']
na_df
stroke_df_copy = stroke_df.copy()
pd.crosstab(stroke_df.loc[stroke_df['bmi'].isna()].gender, stroke_df.loc[stroke_df['bmi'].isna()].stroke, margins=True)
pd.crosstab(stroke_df.loc[stroke_df['bmi'].isna()].smoking_status, stroke_df.loc[stroke_df['bmi'].isna()].stroke, margins=True)
pd.crosstab(stroke_df.loc[stroke_df['bmi'].isna()].hypertension, stroke_df.loc[stroke_df['bmi'].isna()].stroke, margins=True) | code |
90130201/cell_12 | [
"text_plain_output_1.png"
] | import pandas as pd
stroke_df = pd.read_csv('../input/stroke-prediction-dataset/healthcare-dataset-stroke-data.csv')
stroke_df.shape
na_count = []
for i in range(0, len(stroke_df.columns)):
na_count.append(stroke_df[stroke_df.columns[i]].isna().sum())
na_df = pd.DataFrame(zip(stroke_df.columns, na_count))
na_df.columns = ['Variable', 'Counts (NA)']
na_df | code |
105174236/cell_13 | [
"text_html_output_1.png"
] | from sklearn.preprocessing import LabelEncoder
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
df.isnull().sum()
le = LabelEncoder()
df['le_region'] = le.fit_transform(df.region)
df['le_smoker'] = le.fit_transform(df.smoker)
df['le_sex'] = le.fit_transform(df.sex)
ml_df = df.loc[:, ('age', 'le_sex', 'bmi', 'children', 'le_region', 'le_smoker', 'charges')]
ml_df.rename(columns={'le_region': 'region', 'le_smoker': 'smoker', 'le_sex': 'sex'}, inplace=True)
ml_df.head() | code |
105174236/cell_9 | [
"text_html_output_1.png"
] | import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
df.describe() | code |
105174236/cell_23 | [
"text_html_output_1.png"
] | from sklearn import linear_model
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
df.isnull().sum()
le = LabelEncoder()
df['le_region'] = le.fit_transform(df.region)
df['le_smoker'] = le.fit_transform(df.smoker)
df['le_sex'] = le.fit_transform(df.sex)
ml_df = df.loc[:, ('age', 'le_sex', 'bmi', 'children', 'le_region', 'le_smoker', 'charges')]
ml_df.rename(columns={'le_region': 'region', 'le_smoker': 'smoker', 'le_sex': 'sex'}, inplace=True)
columns = ml_df.columns
X = ml_df[columns[:-1]]
y = ml_df[columns[-1]]
X_train, X_test, y_train, y_test = train_test_split(X, y, train_size=0.8)
lin_regr = linear_model.LinearRegression()
lin_regr.fit(X_train, y_train)
lw = lin_regr.coef_
lb = lin_regr.intercept_
ridge = linear_model.Ridge(alpha=0.5)
ridge.fit(X_train, y_train)
rw = ridge.coef_
rb = ridge.intercept_
print(f'The coefficient vector W is: {rw}')
print(f'The intercept b is: {rb}') | code |
105174236/cell_20 | [
"text_html_output_1.png"
] | from sklearn import linear_model
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
df.isnull().sum()
le = LabelEncoder()
df['le_region'] = le.fit_transform(df.region)
df['le_smoker'] = le.fit_transform(df.smoker)
df['le_sex'] = le.fit_transform(df.sex)
ml_df = df.loc[:, ('age', 'le_sex', 'bmi', 'children', 'le_region', 'le_smoker', 'charges')]
ml_df.rename(columns={'le_region': 'region', 'le_smoker': 'smoker', 'le_sex': 'sex'}, inplace=True)
columns = ml_df.columns
X = ml_df[columns[:-1]]
y = ml_df[columns[-1]]
X_train, X_test, y_train, y_test = train_test_split(X, y, train_size=0.8)
lin_regr = linear_model.LinearRegression()
lin_regr.fit(X_train, y_train)
lw = lin_regr.coef_
lb = lin_regr.intercept_
print(f'The coefficient vector W is: {lw}')
print(f'The intercept b is: {lb}') | code |
105174236/cell_6 | [
"text_plain_output_1.png"
] | import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
df.head() | code |
105174236/cell_11 | [
"text_plain_output_1.png"
] | from sklearn.preprocessing import LabelEncoder
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
df.isnull().sum()
le = LabelEncoder()
df['le_region'] = le.fit_transform(df.region)
df['le_smoker'] = le.fit_transform(df.smoker)
df['le_sex'] = le.fit_transform(df.sex)
df.head() | code |
105174236/cell_1 | [
"text_plain_output_1.png"
] | import os
import numpy as np
import pandas as pd
import os
for dirname, _, filenames in os.walk('/kaggle/input'):
for filename in filenames:
print(os.path.join(dirname, filename)) | code |
105174236/cell_7 | [
"text_html_output_1.png"
] | import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
print(df.columns) | code |
105174236/cell_18 | [
"text_plain_output_1.png"
] | from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
df.isnull().sum()
le = LabelEncoder()
df['le_region'] = le.fit_transform(df.region)
df['le_smoker'] = le.fit_transform(df.smoker)
df['le_sex'] = le.fit_transform(df.sex)
ml_df = df.loc[:, ('age', 'le_sex', 'bmi', 'children', 'le_region', 'le_smoker', 'charges')]
ml_df.rename(columns={'le_region': 'region', 'le_smoker': 'smoker', 'le_sex': 'sex'}, inplace=True)
columns = ml_df.columns
X = ml_df[columns[:-1]]
y = ml_df[columns[-1]]
X_train, X_test, y_train, y_test = train_test_split(X, y, train_size=0.8)
print(f'The size of training set is: {X_train.shape}x{y_train.shape}')
print(f'The size of testing set is: {X_test.shape}x{y_test.shape}') | code |
105174236/cell_28 | [
"text_plain_output_1.png"
] | from sklearn import linear_model
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
import numpy as np
import numpy as np # linear algebra
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
df.isnull().sum()
le = LabelEncoder()
df['le_region'] = le.fit_transform(df.region)
df['le_smoker'] = le.fit_transform(df.smoker)
df['le_sex'] = le.fit_transform(df.sex)
ml_df = df.loc[:, ('age', 'le_sex', 'bmi', 'children', 'le_region', 'le_smoker', 'charges')]
ml_df.rename(columns={'le_region': 'region', 'le_smoker': 'smoker', 'le_sex': 'sex'}, inplace=True)
columns = ml_df.columns
X = ml_df[columns[:-1]]
y = ml_df[columns[-1]]
X_train, X_test, y_train, y_test = train_test_split(X, y, train_size=0.8)
lin_regr = linear_model.LinearRegression()
lin_regr.fit(X_train, y_train)
lw = lin_regr.coef_
lb = lin_regr.intercept_
ridge = linear_model.Ridge(alpha=0.5)
ridge.fit(X_train, y_train)
rw = ridge.coef_
rb = ridge.intercept_
lw = np.array(lw).reshape(lw.shape[0], 1).reshape(-1)
rw = np.array(rw).reshape(rw.shape[0], 1).reshape(-1)
coefs = pd.DataFrame(data=np.array([lw, rw, [lb] * lw.shape[0], [rb] * lw.shape[0]]).T, columns=['Linear_W', 'Ridge_W', 'Linear_b', 'Ridge_b'])
coefs.describe() | code |
105174236/cell_8 | [
"text_html_output_1.png"
] | import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
df.info() | code |
105174236/cell_15 | [
"text_plain_output_2.png",
"text_plain_output_1.png"
] | from sklearn.preprocessing import LabelEncoder
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
df.isnull().sum()
le = LabelEncoder()
df['le_region'] = le.fit_transform(df.region)
df['le_smoker'] = le.fit_transform(df.smoker)
df['le_sex'] = le.fit_transform(df.sex)
ml_df = df.loc[:, ('age', 'le_sex', 'bmi', 'children', 'le_region', 'le_smoker', 'charges')]
ml_df.rename(columns={'le_region': 'region', 'le_smoker': 'smoker', 'le_sex': 'sex'}, inplace=True)
ml_df.head() | code |
105174236/cell_16 | [
"text_html_output_1.png"
] | from sklearn.preprocessing import LabelEncoder
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
df.isnull().sum()
le = LabelEncoder()
df['le_region'] = le.fit_transform(df.region)
df['le_smoker'] = le.fit_transform(df.smoker)
df['le_sex'] = le.fit_transform(df.sex)
ml_df = df.loc[:, ('age', 'le_sex', 'bmi', 'children', 'le_region', 'le_smoker', 'charges')]
ml_df.rename(columns={'le_region': 'region', 'le_smoker': 'smoker', 'le_sex': 'sex'}, inplace=True)
columns = ml_df.columns
X = ml_df[columns[:-1]]
y = ml_df[columns[-1]]
print(f'The size of input is: {X.shape}')
print(f'The size of target is: {y.shape}') | code |
105174236/cell_10 | [
"text_html_output_1.png"
] | import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
print('Is there any missing value? \n')
df.isnull().sum() | code |
105174236/cell_27 | [
"text_plain_output_1.png"
] | from sklearn import linear_model
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
import numpy as np
import numpy as np # linear algebra
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
df.isnull().sum()
le = LabelEncoder()
df['le_region'] = le.fit_transform(df.region)
df['le_smoker'] = le.fit_transform(df.smoker)
df['le_sex'] = le.fit_transform(df.sex)
ml_df = df.loc[:, ('age', 'le_sex', 'bmi', 'children', 'le_region', 'le_smoker', 'charges')]
ml_df.rename(columns={'le_region': 'region', 'le_smoker': 'smoker', 'le_sex': 'sex'}, inplace=True)
columns = ml_df.columns
X = ml_df[columns[:-1]]
y = ml_df[columns[-1]]
X_train, X_test, y_train, y_test = train_test_split(X, y, train_size=0.8)
lin_regr = linear_model.LinearRegression()
lin_regr.fit(X_train, y_train)
lw = lin_regr.coef_
lb = lin_regr.intercept_
ridge = linear_model.Ridge(alpha=0.5)
ridge.fit(X_train, y_train)
rw = ridge.coef_
rb = ridge.intercept_
lw = np.array(lw).reshape(lw.shape[0], 1).reshape(-1)
rw = np.array(rw).reshape(rw.shape[0], 1).reshape(-1)
coefs = pd.DataFrame(data=np.array([lw, rw, [lb] * lw.shape[0], [rb] * lw.shape[0]]).T, columns=['Linear_W', 'Ridge_W', 'Linear_b', 'Ridge_b'])
coefs.head() | code |
105174236/cell_12 | [
"text_plain_output_1.png"
] | from sklearn.preprocessing import LabelEncoder
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
df = pd.read_csv('../input/insurance/insurance.csv')
df.isnull().sum()
le = LabelEncoder()
df['le_region'] = le.fit_transform(df.region)
df['le_smoker'] = le.fit_transform(df.smoker)
df['le_sex'] = le.fit_transform(df.sex)
print(f'Before encoding: \n{df.smoker.value_counts()}')
print(f'After encoding: \n{df.le_smoker.value_counts()}') | code |
73097374/cell_9 | [
"text_plain_output_1.png"
] | from sklearn.compose import ColumnTransformer
from sklearn.impute import SimpleImputer
from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import OneHotEncoder
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
train = pd.read_csv('../input/30-days-of-ml/train.csv', index_col=0)
test = pd.read_csv('../input/30-days-of-ml/test.csv', index_col=0)
y = train['target']
features = train.drop(['target'], axis=1)
object_cols = [col for col in features.columns if 'cat' in col]
num_cols = [col for col in features.columns if features[col].dtype in ('int64', 'float64')]
X = features.copy()
X_test = test[features.columns].copy()
numerical_transformer = SimpleImputer(strategy='constant')
categorical_transformer = Pipeline(steps=[('imp', SimpleImputer(strategy='most_frequent')), ('OHen', OneHotEncoder(handle_unknown='ignore', sparse=False))])
processor = ColumnTransformer(transformers=[('num', numerical_transformer, num_cols), ('cat', categorical_transformer, object_cols)])
my_pipeline = Pipeline(steps=[('processor', processor)])
X_pre = pd.DataFrame(my_pipeline.fit_transform(X))
X_pre.index = X.index
X_test_pre = pd.DataFrame(my_pipeline.transform(X_test))
X_test_pre.index = X_test.index
X_train, X_valid, y_train, y_valid = train_test_split(X_pre, y, train_size=0.8, test_size=0.2, random_state=0)
print(X_train, '\n', X_valid) | code |
73097374/cell_4 | [
"text_plain_output_1.png"
] | import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
train = pd.read_csv('../input/30-days-of-ml/train.csv', index_col=0)
test = pd.read_csv('../input/30-days-of-ml/test.csv', index_col=0)
y = train['target']
features = train.drop(['target'], axis=1)
print(features.nunique())
features.head() | code |
73097374/cell_11 | [
"text_html_output_1.png",
"text_plain_output_1.png"
] | from sklearn.compose import ColumnTransformer
from sklearn.impute import SimpleImputer
from sklearn.metrics import mean_squared_error
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import OneHotEncoder
from xgboost import XGBRegressor
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
train = pd.read_csv('../input/30-days-of-ml/train.csv', index_col=0)
test = pd.read_csv('../input/30-days-of-ml/test.csv', index_col=0)
y = train['target']
features = train.drop(['target'], axis=1)
object_cols = [col for col in features.columns if 'cat' in col]
num_cols = [col for col in features.columns if features[col].dtype in ('int64', 'float64')]
X = features.copy()
X_test = test[features.columns].copy()
numerical_transformer = SimpleImputer(strategy='constant')
categorical_transformer = Pipeline(steps=[('imp', SimpleImputer(strategy='most_frequent')), ('OHen', OneHotEncoder(handle_unknown='ignore', sparse=False))])
processor = ColumnTransformer(transformers=[('num', numerical_transformer, num_cols), ('cat', categorical_transformer, object_cols)])
my_pipeline = Pipeline(steps=[('processor', processor)])
X_pre = pd.DataFrame(my_pipeline.fit_transform(X))
X_pre.index = X.index
X_test_pre = pd.DataFrame(my_pipeline.transform(X_test))
X_test_pre.index = X_test.index
model = XGBRegressor(n_estimators=5000, n_jobs=4, learning_rate=0.005, max_depth=5, colsample_bytree=0.5, tree_method='hist', random_state=0)
model.fit(X_train, y_train, early_stopping_rounds=5, eval_set=[(X_valid, y_valid)], eval_metric='rmse', verbose=False)
preds_valid = model.predict(X_valid)
predictions = model.predict(X_test_pre)
print(predictions)
output = pd.DataFrame({'Id': X_test_pre.index, 'target': predictions})
output.to_csv('submission.csv', index=False) | code |
73097374/cell_1 | [
"text_plain_output_1.png"
] | import os
import numpy as np
import pandas as pd
import xgboost as xgb
import os
for dirname, _, filenames in os.walk('/kaggle/input'):
for filename in filenames:
print(os.path.join(dirname, filename)) | code |
73097374/cell_7 | [
"text_plain_output_1.png"
] | import matplotlib.pyplot as plt
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import seaborn as sns
train = pd.read_csv('../input/30-days-of-ml/train.csv', index_col=0)
test = pd.read_csv('../input/30-days-of-ml/test.csv', index_col=0)
y = train['target']
features = train.drop(['target'], axis=1)
object_cols = [col for col in features.columns if 'cat' in col]
num_cols = [col for col in features.columns if features[col].dtype in ('int64', 'float64')]
X = features.copy()
X_test = test[features.columns].copy()
def remove_outlier(df):
Q1 = df.quantile(0.25)
Q3 = df.quantile(0.75)
IQ_range = Q3 - Q1
df_removed = df[~((df < Q1 - 1.5 * IQ_range) | (df > Q3 + 1.5 * IQ_range))]
return df_removed
X[num_cols] = remove_outlier(X[num_cols])
X_test[num_cols] = remove_outlier(X_test[num_cols])
plt.figure(figsize=(10, 8))
sns.boxplot(data=X[num_cols]).set(title='Review X outlier-removed data')
plt.figure(figsize=(10, 8))
sns.boxplot(data=X_test[num_cols]).set(title='Review X_test outlier-removed data') | code |
73097374/cell_8 | [
"text_plain_output_1.png"
] | from sklearn.compose import ColumnTransformer
from sklearn.impute import SimpleImputer
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import OneHotEncoder
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
train = pd.read_csv('../input/30-days-of-ml/train.csv', index_col=0)
test = pd.read_csv('../input/30-days-of-ml/test.csv', index_col=0)
y = train['target']
features = train.drop(['target'], axis=1)
object_cols = [col for col in features.columns if 'cat' in col]
num_cols = [col for col in features.columns if features[col].dtype in ('int64', 'float64')]
X = features.copy()
X_test = test[features.columns].copy()
numerical_transformer = SimpleImputer(strategy='constant')
categorical_transformer = Pipeline(steps=[('imp', SimpleImputer(strategy='most_frequent')), ('OHen', OneHotEncoder(handle_unknown='ignore', sparse=False))])
processor = ColumnTransformer(transformers=[('num', numerical_transformer, num_cols), ('cat', categorical_transformer, object_cols)])
my_pipeline = Pipeline(steps=[('processor', processor)])
X_pre = pd.DataFrame(my_pipeline.fit_transform(X))
X_pre.index = X.index
X_test_pre = pd.DataFrame(my_pipeline.transform(X_test))
X_test_pre.index = X_test.index
print(X_pre, '\n', X_test_pre) | code |
73097374/cell_3 | [
"text_plain_output_1.png",
"image_output_2.png",
"image_output_1.png"
] | import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
train = pd.read_csv('../input/30-days-of-ml/train.csv', index_col=0)
test = pd.read_csv('../input/30-days-of-ml/test.csv', index_col=0)
train.head()
print(train) | code |
73097374/cell_10 | [
"text_plain_output_1.png"
] | from sklearn.metrics import mean_squared_error
from xgboost import XGBRegressor
model = XGBRegressor(n_estimators=5000, n_jobs=4, learning_rate=0.005, max_depth=5, colsample_bytree=0.5, tree_method='hist', random_state=0)
model.fit(X_train, y_train, early_stopping_rounds=5, eval_set=[(X_valid, y_valid)], eval_metric='rmse', verbose=False)
preds_valid = model.predict(X_valid)
print(mean_squared_error(y_valid, preds_valid, squared=False)) | code |
33111929/cell_9 | [
"text_plain_output_1.png",
"image_output_1.png"
] | import pandas as pd
data = pd.read_csv('../input/diamonds/diamonds.csv')
data.dtypes
data.head(10) | code |
33111929/cell_19 | [
"text_html_output_1.png"
] | import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns
data = pd.read_csv('../input/diamonds/diamonds.csv')
data.dtypes
data = data.drop(data.loc[data.x <= 0].index)
data = data.drop(data.loc[data.y <= 0].index)
data = data.drop(data.loc[data.z <= 0].index)
data['ratio'] = data.x / data.y
premium = ['D', 'E', 'F', 'G', 'H']
def data_split(status):
if status in premium:
return 'premium'
else:
return 'normal'
def data_split_num(status):
if status in premium:
return 1
else:
return 0
data['data_split'] = data['color'].apply(data_split)
data['data_split_num'] = data['color'].apply(data_split_num)
#correlation matrix for 15 variables with largest correlation
corrmat = data.corr()
f, ax = plt.subplots(figsize=(12, 9))
k = 8 #number of variables for heatmap
cols = corrmat.nlargest(k, 'price')['price'].index
cm = np.corrcoef(data[cols].values.T)
# Generate a mask for the upper triangle
mask = np.zeros_like(cm, dtype=np.bool)
mask[np.triu_indices_from(mask)] = True
hm = sns.heatmap(cm, vmax=1, mask=mask, cbar=True, annot=True, square=True, fmt='.2f', annot_kws={'size': 10}, yticklabels=cols.values, xticklabels=cols.values)
plt.show()
sns.countplot(y=data.cut)
plt.show() | code |
33111929/cell_1 | [
"text_plain_output_1.png"
] | import os
import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report, confusion_matrix
from sklearn.model_selection import cross_val_score
from sklearn.metrics import accuracy_score
from sklearn.ensemble import RandomForestClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.naive_bayes import GaussianNB
from sklearn.linear_model import Perceptron
from sklearn.linear_model import SGDClassifier
from sklearn.tree import DecisionTreeClassifier
import os
for dirname, _, filenames in os.walk('/kaggle/input'):
for filename in filenames:
print(os.path.join(dirname, filename)) | code |
33111929/cell_7 | [
"text_plain_output_1.png",
"image_output_1.png"
] | import pandas as pd
data = pd.read_csv('../input/diamonds/diamonds.csv')
data.dtypes | code |
33111929/cell_8 | [
"image_output_1.png"
] | import pandas as pd
data = pd.read_csv('../input/diamonds/diamonds.csv')
data.dtypes
data.info() | code |
33111929/cell_15 | [
"text_html_output_1.png"
] | import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns
data = pd.read_csv('../input/diamonds/diamonds.csv')
data.dtypes
data = data.drop(data.loc[data.x <= 0].index)
data = data.drop(data.loc[data.y <= 0].index)
data = data.drop(data.loc[data.z <= 0].index)
data['ratio'] = data.x / data.y
premium = ['D', 'E', 'F', 'G', 'H']
def data_split(status):
if status in premium:
return 'premium'
else:
return 'normal'
def data_split_num(status):
if status in premium:
return 1
else:
return 0
data['data_split'] = data['color'].apply(data_split)
data['data_split_num'] = data['color'].apply(data_split_num)
corrmat = data.corr()
f, ax = plt.subplots(figsize=(12, 9))
k = 8
cols = corrmat.nlargest(k, 'price')['price'].index
cm = np.corrcoef(data[cols].values.T)
mask = np.zeros_like(cm, dtype=np.bool)
mask[np.triu_indices_from(mask)] = True
hm = sns.heatmap(cm, vmax=1, mask=mask, cbar=True, annot=True, square=True, fmt='.2f', annot_kws={'size': 10}, yticklabels=cols.values, xticklabels=cols.values)
plt.show() | code |
33111929/cell_17 | [
"text_html_output_1.png"
] | import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import seaborn as sns
data = pd.read_csv('../input/diamonds/diamonds.csv')
data.dtypes
data = data.drop(data.loc[data.x <= 0].index)
data = data.drop(data.loc[data.y <= 0].index)
data = data.drop(data.loc[data.z <= 0].index)
data['ratio'] = data.x / data.y
premium = ['D', 'E', 'F', 'G', 'H']
def data_split(status):
if status in premium:
return 'premium'
else:
return 'normal'
def data_split_num(status):
if status in premium:
return 1
else:
return 0
data['data_split'] = data['color'].apply(data_split)
data['data_split_num'] = data['color'].apply(data_split_num)
#correlation matrix for 15 variables with largest correlation
corrmat = data.corr()
f, ax = plt.subplots(figsize=(12, 9))
k = 8 #number of variables for heatmap
cols = corrmat.nlargest(k, 'price')['price'].index
cm = np.corrcoef(data[cols].values.T)
# Generate a mask for the upper triangle
mask = np.zeros_like(cm, dtype=np.bool)
mask[np.triu_indices_from(mask)] = True
hm = sns.heatmap(cm, vmax=1, mask=mask, cbar=True, annot=True, square=True, fmt='.2f', annot_kws={'size': 10}, yticklabels=cols.values, xticklabels=cols.values)
plt.show()
print(' Diamond Carat = ' + str(np.mean(data.carat)))
plt.subplots(figsize=(10, 7))
sns.distplot(data.carat)
plt.show() | code |
33111929/cell_14 | [
"text_plain_output_1.png"
] | import pandas as pd
data = pd.read_csv('../input/diamonds/diamonds.csv')
data.dtypes
data = data.drop(data.loc[data.x <= 0].index)
data = data.drop(data.loc[data.y <= 0].index)
data = data.drop(data.loc[data.z <= 0].index)
data['ratio'] = data.x / data.y
premium = ['D', 'E', 'F', 'G', 'H']
def data_split(status):
if status in premium:
return 'premium'
else:
return 'normal'
def data_split_num(status):
if status in premium:
return 1
else:
return 0
data['data_split'] = data['color'].apply(data_split)
data['data_split_num'] = data['color'].apply(data_split_num)
data.head() | code |
33111929/cell_10 | [
"text_plain_output_1.png"
] | import pandas as pd
data = pd.read_csv('../input/diamonds/diamonds.csv')
data.dtypes
data.describe() | code |
17144077/cell_2 | [
"text_html_output_1.png"
] | import os
import os
import numpy as np
import pandas as pd
import os
import os
import cv2
import random
import numpy as np
import pandas as pd
import scipy as sp
import torch
from fastai.vision import *
import glob
print(os.listdir('../input/fastai-pretrained-models')) | code |
17144077/cell_1 | [
"text_plain_output_1.png"
] | import os
import numpy as np
import pandas as pd
import os
print(os.listdir('../input')) | code |
17144077/cell_8 | [
"text_html_output_1.png"
] | import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
SIZE = 224
train_df = pd.read_csv(PATH + '/train.csv')
test_df = pd.read_csv(PATH + '/sample_submission.csv')
train = ImageList.from_df(train_df, path=PATH, cols='id_code', folder='train_images', suffix='.png')
test = ImageList.from_df(test_df, path=PATH, cols='id_code', folder='test_images', suffix='.png')
from sklearn.metrics import cohen_kappa_score
data = train.split_by_rand_pct(0.2).label_from_df(cols='diagnosis', label_cls=FloatList).add_test(test).transform(get_transforms(), size=SIZE).databunch(path=Path('.'), bs=32).normalize(imagenet_stats)
data.show_batch(rows=3, figsize=(7, 6)) | code |
17144077/cell_15 | [
"text_html_output_1.png",
"image_output_2.png",
"image_output_1.png"
] | from sklearn.metrics import cohen_kappa_score
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import torch
SIZE = 224
train_df = pd.read_csv(PATH + '/train.csv')
test_df = pd.read_csv(PATH + '/sample_submission.csv')
train = ImageList.from_df(train_df, path=PATH, cols='id_code', folder='train_images', suffix='.png')
test = ImageList.from_df(test_df, path=PATH, cols='id_code', folder='test_images', suffix='.png')
def quadratic_kappa(y_hat, y):
return torch.tensor(cohen_kappa_score(torch.round(y_hat), y, weights='quadratic'), device='cuda:0')
from sklearn.metrics import cohen_kappa_score
data = train.split_by_rand_pct(0.2).label_from_df(cols='diagnosis', label_cls=FloatList).add_test(test).transform(get_transforms(), size=SIZE).databunch(path=Path('.'), bs=32).normalize(imagenet_stats)
learn = cnn_learner(data, models.resnet101, metrics=[quadratic_kappa], pretrained=True)
learn.fit_one_cycle(6, slice(0.00275, 0.0275))
learn.save('stage1')
learn.unfreeze()
learn.lr_find()
lr = 0.00275
learn.fit_one_cycle(4, slice(1.5e-06, lr / 8), wd=0.05) | code |
17144077/cell_16 | [
"text_plain_output_1.png",
"image_output_1.png"
] | from sklearn.metrics import cohen_kappa_score
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import torch
SIZE = 224
train_df = pd.read_csv(PATH + '/train.csv')
test_df = pd.read_csv(PATH + '/sample_submission.csv')
train = ImageList.from_df(train_df, path=PATH, cols='id_code', folder='train_images', suffix='.png')
test = ImageList.from_df(test_df, path=PATH, cols='id_code', folder='test_images', suffix='.png')
def quadratic_kappa(y_hat, y):
return torch.tensor(cohen_kappa_score(torch.round(y_hat), y, weights='quadratic'), device='cuda:0')
from sklearn.metrics import cohen_kappa_score
data = train.split_by_rand_pct(0.2).label_from_df(cols='diagnosis', label_cls=FloatList).add_test(test).transform(get_transforms(), size=SIZE).databunch(path=Path('.'), bs=32).normalize(imagenet_stats)
learn = cnn_learner(data, models.resnet101, metrics=[quadratic_kappa], pretrained=True)
learn.fit_one_cycle(6, slice(0.00275, 0.0275))
learn.save('stage1')
learn.unfreeze()
learn.lr_find()
lr = 0.00275
learn.fit_one_cycle(4, slice(1.5e-06, lr / 8), wd=0.05)
valid_preds, valid_y = learn.TTA(ds_type=DatasetType.Valid)
test_preds, _ = learn.TTA(ds_type=DatasetType.Test) | code |
17144077/cell_14 | [
"image_output_1.png"
] | from sklearn.metrics import cohen_kappa_score
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import torch
SIZE = 224
train_df = pd.read_csv(PATH + '/train.csv')
test_df = pd.read_csv(PATH + '/sample_submission.csv')
train = ImageList.from_df(train_df, path=PATH, cols='id_code', folder='train_images', suffix='.png')
test = ImageList.from_df(test_df, path=PATH, cols='id_code', folder='test_images', suffix='.png')
def quadratic_kappa(y_hat, y):
return torch.tensor(cohen_kappa_score(torch.round(y_hat), y, weights='quadratic'), device='cuda:0')
from sklearn.metrics import cohen_kappa_score
data = train.split_by_rand_pct(0.2).label_from_df(cols='diagnosis', label_cls=FloatList).add_test(test).transform(get_transforms(), size=SIZE).databunch(path=Path('.'), bs=32).normalize(imagenet_stats)
learn = cnn_learner(data, models.resnet101, metrics=[quadratic_kappa], pretrained=True)
learn.fit_one_cycle(6, slice(0.00275, 0.0275))
learn.save('stage1')
learn.unfreeze()
learn.lr_find()
learn.recorder.plot(suggestion=True) | code |
17144077/cell_22 | [
"text_plain_output_1.png"
] | from sklearn.metrics import cohen_kappa_score
import cv2
import numpy as np
import numpy as np # linear algebra
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import scipy as sp
SIZE = 224
train_df = pd.read_csv(PATH + '/train.csv')
test_df = pd.read_csv(PATH + '/sample_submission.csv')
def crop_image(img, tol=7):
mask = img > tol
return img[np.ix_(mask.any(1), mask.any(0))]
def open_aptos2019_image(fn, convert_mode, after_open) -> Image:
image = cv2.imread(fn)
image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
image = crop_image(image)
image = cv2.resize(image, (SIZE, SIZE))
image = cv2.addWeighted(image, 4, cv2.GaussianBlur(image, (0, 0), SIZE / 10), -4, 128)
return Image(pil2tensor(image, np.float32).div_(255))
vision.data.open_image = open_aptos2019_image
class OptimizedRounder(object):
def __init__(self):
self.coef_ = 0
def _kappa_loss(self, coef, X, y):
X_p = np.copy(X)
for i, pred in enumerate(X_p):
if pred < coef[0]:
X_p[i] = 0
elif pred >= coef[0] and pred < coef[1]:
X_p[i] = 1
elif pred >= coef[1] and pred < coef[2]:
X_p[i] = 2
elif pred >= coef[2] and pred < coef[3]:
X_p[i] = 3
else:
X_p[i] = 4
ll = cohen_kappa_score(y, X_p, weights='quadratic')
return -ll
def fit(self, X, y):
loss_partial = partial(self._kappa_loss, X=X, y=y)
initial_coef = [0.5, 1.5, 2.5, 3.5]
self.coef_ = sp.optimize.minimize(loss_partial, initial_coef, method='nelder-mead')
def predict(self, X, coef):
X_p = np.copy(X)
for i, pred in enumerate(X_p):
if pred < coef[0]:
X_p[i] = 0
elif pred >= coef[0] and pred < coef[1]:
X_p[i] = 1
elif pred >= coef[1] and pred < coef[2]:
X_p[i] = 2
elif pred >= coef[2] and pred < coef[3]:
X_p[i] = 3
else:
X_p[i] = 4
return X_p
def coefficients(self):
return self.coef_['x']
optR = OptimizedRounder()
optR.fit(valid_preds, valid_y)
coefficients = optR.coefficients()
valid_predictions = optR.predict(valid_preds, coefficients)[:, 0].astype(int)
test_predictions = optR.predict(test_preds, coefficients)[:, 0].astype(int)
valid_score = cohen_kappa_score(valid_y.numpy().astype(int), valid_predictions, weights='quadratic')
test_df.diagnosis = test_predictions
test_df.to_csv('submission.csv', index=None)
test_df.head() | code |
17144077/cell_12 | [
"text_plain_output_1.png"
] | from sklearn.metrics import cohen_kappa_score
import pandas as pd
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import torch
SIZE = 224
train_df = pd.read_csv(PATH + '/train.csv')
test_df = pd.read_csv(PATH + '/sample_submission.csv')
train = ImageList.from_df(train_df, path=PATH, cols='id_code', folder='train_images', suffix='.png')
test = ImageList.from_df(test_df, path=PATH, cols='id_code', folder='test_images', suffix='.png')
def quadratic_kappa(y_hat, y):
return torch.tensor(cohen_kappa_score(torch.round(y_hat), y, weights='quadratic'), device='cuda:0')
from sklearn.metrics import cohen_kappa_score
data = train.split_by_rand_pct(0.2).label_from_df(cols='diagnosis', label_cls=FloatList).add_test(test).transform(get_transforms(), size=SIZE).databunch(path=Path('.'), bs=32).normalize(imagenet_stats)
learn = cnn_learner(data, models.resnet101, metrics=[quadratic_kappa], pretrained=True)
learn.fit_one_cycle(6, slice(0.00275, 0.0275))
learn.recorder.plot_losses()
learn.recorder.plot_metrics() | code |
105190429/cell_9 | [
"image_output_1.png"
] | import pandas as pd
data = pd.read_csv('../input/heart-attack-analysis-prediction-dataset/heart.csv')
data.head() | code |
105190429/cell_29 | [
"image_output_1.png"
] | import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns
data = pd.read_csv('../input/heart-attack-analysis-prediction-dataset/heart.csv')
data.describe().T
data.isnull().sum()
duplicate = data[data.duplicated()]
data.drop_duplicates(inplace=True)
data = data.reset_index(drop=True)
plt.figure(figsize=(12,6))
sns.set(style='darkgrid')
ax = sns.countplot(y = 'output', data=data, palette='viridis')
ax.set_yticklabels(['Less Chance','Higher Chance'])
plt.ylabel('Chance to Have Heart Attack')
plt.xlabel('Number of Person')
plt.title('Output distribution', fontsize=20, loc = 'center')
plt.show()
fig, ax = plt.subplots(1, 2, figsize=(15, 8))
sns.histplot(x='age', data=data, kde=True, hue='output', ax=ax[0], palette='viridis')
ax[0].set_title('Age distribution on Output')
ax[0].set_ylabel('Number of Person')
ax[0].set_xlabel('age')
ax[0].legend(title='Heart Attack Chances', labels=['Higher Chance', 'Less Chance'], loc='upper right')
sns.boxplot(x='output', data=data, y='age', ax=ax[1], palette='viridis')
ax[1].set_title('Age distribution on Output')
ax[1].set_xticklabels(['Less Chance', 'Higher Chance'])
ax[1].set_xlabel('Chance to Have Heart Attack')
plt.tight_layout()
plt.show() | code |
105190429/cell_26 | [
"text_plain_output_1.png"
] | import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns
data = pd.read_csv('../input/heart-attack-analysis-prediction-dataset/heart.csv')
data.describe().T
data.isnull().sum()
duplicate = data[data.duplicated()]
data.drop_duplicates(inplace=True)
data = data.reset_index(drop=True)
plt.figure(figsize=(12, 6))
sns.set(style='darkgrid')
ax = sns.countplot(y='output', data=data, palette='viridis')
ax.set_yticklabels(['Less Chance', 'Higher Chance'])
plt.ylabel('Chance to Have Heart Attack')
plt.xlabel('Number of Person')
plt.title('Output distribution', fontsize=20, loc='center')
plt.show() | code |
105190429/cell_11 | [
"text_html_output_1.png"
] | import pandas as pd
data = pd.read_csv('../input/heart-attack-analysis-prediction-dataset/heart.csv')
data.describe().T
data.info() | code |
105190429/cell_1 | [
"text_plain_output_1.png"
] | !pip install catboost
!pip install shap | code |
105190429/cell_18 | [
"text_plain_output_1.png"
] | import pandas as pd
data = pd.read_csv('../input/heart-attack-analysis-prediction-dataset/heart.csv')
data.describe().T
data.isnull().sum()
duplicate = data[data.duplicated()]
duplicate | code |
105190429/cell_32 | [
"image_output_1.png"
] | import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns
data = pd.read_csv('../input/heart-attack-analysis-prediction-dataset/heart.csv')
data.describe().T
data.isnull().sum()
duplicate = data[data.duplicated()]
data.drop_duplicates(inplace=True)
data = data.reset_index(drop=True)
plt.figure(figsize=(12,6))
sns.set(style='darkgrid')
ax = sns.countplot(y = 'output', data=data, palette='viridis')
ax.set_yticklabels(['Less Chance','Higher Chance'])
plt.ylabel('Chance to Have Heart Attack')
plt.xlabel('Number of Person')
plt.title('Output distribution', fontsize=20, loc = 'center')
plt.show()
fig, ax = plt.subplots(1,2, figsize=(15,8))
sns.histplot(x='age', data=data, kde=True, hue='output',
ax=ax[0], palette = 'viridis')
ax[0].set_title('Age distribution on Output')
ax[0].set_ylabel('Number of Person')
ax[0].set_xlabel('age')
ax[0].legend(title='Heart Attack Chances',
labels=['Higher Chance','Less Chance'],
loc='upper right')
sns.boxplot(x='output', data=data, y='age', ax=ax[1], palette = 'viridis')
ax[1].set_title('Age distribution on Output')
ax[1].set_xticklabels(['Less Chance','Higher Chance'])
ax[1].set_xlabel('Chance to Have Heart Attack')
plt.tight_layout()
plt.show()
fig, ax = plt.subplots(1, 2, figsize=(12, 6))
sns.countplot(x='sex', data=data, hue='output', palette='viridis', ax=ax[0])
ax[0].set_title('Sex distribution on output', fontsize=15)
ax[0].legend(title='Heart Attack Chances', labels=['Less Chance', 'Higher Chance'], loc='upper left')
ax[0].set_ylabel('Number of Person')
sns.countplot(x='sex', data=data, palette='flare', ax=ax[1])
ax[1].set_title('Sex distribution', fontsize=15)
ax[1].set_ylabel('Number of Person')
plt.tight_layout()
plt.show() | code |
105190429/cell_35 | [
"image_output_1.png"
] | import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns
data = pd.read_csv('../input/heart-attack-analysis-prediction-dataset/heart.csv')
data.describe().T
data.isnull().sum()
duplicate = data[data.duplicated()]
data.drop_duplicates(inplace=True)
data = data.reset_index(drop=True)
plt.figure(figsize=(12,6))
sns.set(style='darkgrid')
ax = sns.countplot(y = 'output', data=data, palette='viridis')
ax.set_yticklabels(['Less Chance','Higher Chance'])
plt.ylabel('Chance to Have Heart Attack')
plt.xlabel('Number of Person')
plt.title('Output distribution', fontsize=20, loc = 'center')
plt.show()
fig, ax = plt.subplots(1,2, figsize=(15,8))
sns.histplot(x='age', data=data, kde=True, hue='output',
ax=ax[0], palette = 'viridis')
ax[0].set_title('Age distribution on Output')
ax[0].set_ylabel('Number of Person')
ax[0].set_xlabel('age')
ax[0].legend(title='Heart Attack Chances',
labels=['Higher Chance','Less Chance'],
loc='upper right')
sns.boxplot(x='output', data=data, y='age', ax=ax[1], palette = 'viridis')
ax[1].set_title('Age distribution on Output')
ax[1].set_xticklabels(['Less Chance','Higher Chance'])
ax[1].set_xlabel('Chance to Have Heart Attack')
plt.tight_layout()
plt.show()
fig, ax = plt.subplots(1,2, figsize=(12,6))
sns.countplot(x='sex', data=data, hue='output', palette='viridis', ax=ax[0])
ax[0].set_title('Sex distribution on output', fontsize=15)
ax[0].legend(title='Heart Attack Chances',
labels=['Less Chance','Higher Chance'],
loc='upper left')
ax[0].set_ylabel('Number of Person')
sns.countplot(x='sex', data=data, palette='flare', ax=ax[1])
ax[1].set_title('Sex distribution', fontsize=15)
ax[1].set_ylabel('Number of Person')
plt.tight_layout()
plt.show()
fig, ax = plt.subplots(1, 2, figsize=(12, 6))
sns.countplot(x='cp', data=data, hue='output', palette='viridis', ax=ax[0])
ax[0].set_title('Chest Pain Type Distribution on Output')
ax[0].set_ylabel('Number of Person')
ax[0].set_xlabel('Chest Pain Type')
ax[0].legend(title='Heart Attack Chances', labels=['Less Chance', 'Higher Chance'], loc='upper right')
ax[0].set_xticklabels(['typical angina', 'atypical agina', 'non-anginal pain', 'asymptomatic'], rotation=-45)
sns.countplot(x='cp', data=data, ax=ax[1], palette='flare')
ax[1].set_title('Chest Pain Type Distribution')
ax[1].set_ylabel('Number of Person')
ax[1].set_xlabel('Chest Pain Type')
ax[1].set_xticklabels(['typical angina', 'atypical agina', 'non-anginal pain', 'asymptomatic'], rotation=-45)
plt.tight_layout()
plt.show() | code |
105190429/cell_24 | [
"text_html_output_1.png"
] | import pandas as pd
data = pd.read_csv('../input/heart-attack-analysis-prediction-dataset/heart.csv')
data.describe().T
data.isnull().sum()
duplicate = data[data.duplicated()]
data.drop_duplicates(inplace=True)
data = data.reset_index(drop=True)
print(f'There are {data.shape[0]} records and {data.shape[1]} columns on this dataset.') | code |
105190429/cell_14 | [
"text_plain_output_1.png"
] | import pandas as pd
data = pd.read_csv('../input/heart-attack-analysis-prediction-dataset/heart.csv')
data.describe().T
data.isnull().sum() | code |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.