app.py

import gradio as gr
import joblib
import pandas as pd
import numpy as np
from sklearn.preprocessing import LabelEncoder, StandardScaler, OneHotEncoder
from sklearn.impute import KNNImputer
from sklearn.decomposition import PCA

# Load your saved model
# model = joblib.load("ann_model.joblib")

# # Define the prediction function
def predict(age, workclass, education, marital_status, occupation, relationship, race, gender, capital_gain, capital_loss, hours_per_week, native_country):
    features = [age, workclass, education, marital_status, occupation, relationship, race, gender, capital_gain, capital_loss, hours_per_week, native_country]
    columns = [
    "age", "workclass", "educational-num", "marital-status", "occupation", 
    "relationship", "race", "gender", "capital-gain", "capital-loss", 
    "hours-per-week", "native-country"]
    df = pd.DataFrame(index=features, columns=columns)
    fixed_features = cleaning_features(df)
    # prediction = model.predict(features)
    # prediction = 1
    # return "Income >50K" if prediction == 1 else "Income <=50K"
    return print(fixed_features)

def cleaning_features(data):
    le = LabelEncoder()
    scaler = StandardScaler()
    encoder = OneHotEncoder(sparse=False)
    numeric_cols = ['age', 'educational-num', 'hours-per-week']
    columns_to_encode = ['race','marital-status','relationship']
    
    
    # 1. Scale numerical features
    data[numeric_cols] = scaler.fit_transform(data[numeric_cols])

    # 2. Label encode gender and income
    data['gender'] = le.fit_transform(data['gender'])
    data['educational-num'] = le.fit_transform(data['educational-num'])
    
    # 3. One-hot encode race
    for N in columns_to_encode:
        race_encoded = encoder.fit_transform(data[[N]])
        race_encoded_cols = encoder.get_feature_names_out([N])
        race_encoded_df = pd.DataFrame(race_encoded, columns=race_encoded_cols, index=data.index)
        # Combine the encoded data with original dataframe
        data = pd.concat([data.drop(N, axis=1), race_encoded_df], axis=1)
    # Binarize native country
    data['native-country'] = data['native-country'].apply(lambda x: x == 'United-States')
    data['native-country'] = data['native-country'].astype(int)
    data = pca(data)
    return data

# def pca(data):
#     encoder = OneHotEncoder(sparse_output=False)
#     one_hot_encoded = encoder.fit_transform(data[['workclass', 'occupation']])
#     encoded_columns_df = pd.DataFrame(one_hot_encoded, columns=encoder.get_feature_names_out())
#     pca_net = PCA(n_components=10)
#     pca_result_net = pca_net.fit_transform(encoded_columns_df)
#     pca_columns = [f'pca_component_{i+1}' for i in range(10)]
#     pca_df = pd.DataFrame(pca_result_net, columns=pca_columns)
#     data = data.drop(columns=['workclass', 'occupation'], axis=1) #remove the original columns
#     data = pd.concat([data, pca_df], axis=1)
#     return data


def pca(data):
    encoder = joblib.load('onehot_encoder.joblib')
    pca_model = joblib.load('pca.joblib')
    one_hot_encoded = encoder.transform(data[['workclass', 'occupation']])
    encoded_columns_df = pd.DataFrame(one_hot_encoded, columns=encoder.get_feature_names_out())
    pca_result_net = pca_model.transform(encoded_columns_df)
    pca_columns = [f'pca_component_{i+1}' for i in range(pca_model.n_components_)]
    pca_df = pd.DataFrame(pca_result_net, columns=pca_columns)
    data = data.drop(columns=['workclass', 'occupation'], axis=1)
    data = pd.concat([data, pca_df], axis=1) 
    return data

def hbdscan_tranform(df_transformed):
    df_transformed['capital-gain'] = np.log1p(df_transformed['capital-gain'])
    df_transformed['capital-loss'] = np.log1p(df_transformed['capital-loss'])
    
    # Apply RobustScaler to all numerical features
    numerical_features = ['age', 'capital-gain', 'capital-loss', 'hours-per-week']
    scaler = RobustScaler()
    df_transformed[numerical_features] = scaler.fit_transform(df_transformed[numerical_features])
    return df_transformed


# Create the Gradio interface
interface = gr.Interface(
    fn=predict,
    inputs=[
        gr.Slider(18, 90, step=1, label="Age"),
        gr.Dropdown(
            ["Private", "Self-emp-not-inc", "Self-emp-inc", "Federal-gov", 
             "Local-gov", "State-gov", "Without-pay", "Never-worked"], 
            label="Workclass"
        ),
        gr.Dropdown(
            ["Bachelors", "Some-college", "11th", "HS-grad", "Prof-school", 
             "Assoc-acdm", "Assoc-voc", "9th", "7th-8th", "12th", "Masters", 
             "1st-4th", "10th", "Doctorate", "5th-6th", "Preschool"], 
            label="Education"
        ),
        gr.Dropdown(
            ["Married-civ-spouse", "Divorced", "Never-married", "Separated", 
             "Widowed", "Married-spouse-absent", "Married-AF-spouse"], 
            label="Marital Status"
        ),
        gr.Dropdown(
            ["Tech-support", "Craft-repair", "Other-service", "Sales", 
             "Exec-managerial", "Prof-specialty", "Handlers-cleaners", 
             "Machine-op-inspct", "Adm-clerical", "Farming-fishing", 
             "Transport-moving", "Priv-house-serv", "Protective-serv", 
             "Armed-Forces"], 
            label="Occupation"
        ),
        gr.Dropdown(
            ["Wife", "Husband", "Own-child", "Unmarried", "Other-relative", "Not-in-family"], 
            label="Relationship"
        ),
        gr.Dropdown(
            ["White", "Black", "Asian-Pac-Islander", "Amer-Indian-Eskimo", "Other"], 
            label="Race"
        ),
        gr.Dropdown(
            ["Male", "Female"], 
            label="Gender"
        ),
        gr.Slider(1, 90, step=1, label="Hours Per Week"),
        gr.Slider(0, 100000, step=100, label="Capital Gain"),
        gr.Slider(0, 5000, step=50, label="Capital Loss"),
        gr.Dropdown(
            ["United-States", "Other"], 
            label="Native Country"
        )
    ],
    outputs="text",
    title="Adult Income Predictor"
)

# Launch the app
interface.launch()