File size: 6,927 Bytes
f23dbc8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
21e5d2f
f23dbc8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
# Import python lib
import streamlit as st
import time
import pandas as pd
import numpy as np
from surprise import Dataset, Reader
from surprise import KNNBaseline

# Import wine dataframes
df_wine_model = pd.read_pickle('./data/df_wine_us_rate.pkl')
df_wine_combi = pd.read_pickle('./data/df_wine_combi.pkl')

# Instantiate the list of wine traits
all_traits = ['almond', 'anise', 'apple', 'apricot', 'baked', 'baking_spices', 'berry', 'black_cherry', 'black_currant', 'black_pepper', 'black_tea', 'blackberry', 'blueberry', 
              'boysenberry', 'bramble', 'bright', 'butter', 'candy', 'caramel', 'cardamom', 'cassis', 'cedar', 'chalk', 'cherry', 'chocolate', 'cinnamon', 'citrus', 'clean', 'closed',
              'clove', 'cocoa', 'coffee', 'cola', 'complex', 'concentrated', 'cranberry', 'cream', 'crisp', 'dark', 'dark_chocolate', 'dense', 'depth', 'dried_herb', 'dry', 'dust',
              'earth', 'edgy', 'elderberry', 'elegant', 'fennel', 'firm', 'flower', 'forest_floor', 'french_oak', 'fresh', 'fruit', 'full_bodied', 'game', 'grapefruit', 'graphite',
              'green', 'gripping', 'grippy', 'hearty', 'herb', 'honey', 'honeysuckle', 'jam', 'juicy', 'lavender', 'leafy', 'lean', 'leather', 'lemon', 'lemon_peel', 'length', 'licorice',
              'light_bodied', 'lime', 'lush', 'meaty', 'medium_bodied', 'melon', 'milk_chocolate', 'minerality', 'mint', 'nutmeg', 'oak', 'olive', 'orange', 'orange_peel', 'peach',
              'pear', 'pencil_lead', 'pepper', 'pine', 'pineapple', 'plum', 'plush', 'polished', 'pomegranate', 'powerful', 'purple', 'purple_flower', 'raspberry', 'refreshing',
              'restrained', 'rich', 'ripe', 'robust', 'rose', 'round', 'sage', 'salt', 'savory', 'sharp', 'silky', 'smoke', 'smoked_meat', 'smooth', 'soft', 'sparkling', 'spice',
              'steel', 'stone', 'strawberry', 'succulent', 'supple', 'sweet', 'tangy', 'tannin', 'tar', 'tart', 'tea', 'thick', 'thyme', 'tight', 'toast', 'tobacco', 'tropical_fruit',
              'vanilla', 'velvety', 'vibrant', 'violet', 'warm', 'weight', 'wet_rocks', 'white', 'white_pepper', 'wood']

#---------------------------------------------------------------------------------------------------------

# Function to instantiate the model & return the est recsys scores
def recommend_scores():
    
    # Instantiate reader & data for surprise
    reader = Reader(rating_scale=(88, 100))
    data = Dataset.load_from_df(df_wine_model, reader)
    
    # Instantiate recsys model
    sim_options={'name':'cosine'}
    model = KNNBaseline(k=35, min_k=1, sim_options=sim_options, verbose=False)

    # Train & fit the data into model
    train=data.build_full_trainset()
    model.fit(train)

    # Start the model to compute the best estimate match score on wine list
    recommend_list = []
    user_wines = df_wine_model[df_wine_model.taster_name == 'mockuser']['title'].unique()
    not_user_wines = []
    
    for wine in df_wine_model['title'].unique():
        if wine not in user_wines:
            not_user_wines.append(wine)

    for wine in not_user_wines:
        wine_compatibility = []
        prediction = model.predict(uid='mockuser', iid=wine)
        wine_compatibility.append(prediction.iid)
        wine_compatibility.append(prediction.est)
        recommend_list.append(wine_compatibility)
        
    result_df = pd.DataFrame(recommend_list, columns = ['title', 'est_match_pts'])
    
    return result_df

def add_bg_from_url():
    st.markdown(
        f"""
        <style>
        
        [data-testid="stAppViewContainer"] {{
        background-image: url("https://www.ackerwines.com/wp-content/uploads/2021/09/Montrose4.jpg");
        background-attachment: fixed;
        background-size: cover      
        }}
        
        [data-testid="stVerticalBlock"] {{
        background-color: rgba(255,255,255,0.75)
        }}
        
        </style>
        """,
        unsafe_allow_html=True
    )

#----------------------------------------------------------------------------------------------------------

st.title("Which wine should I get?")
st.text("")
st.write("You can type the wine traits that you want in the dropdown list below")
add_bg_from_url()

select_temptrait = st.multiselect('Choose the traits that you want in your wine', options = all_traits)

if st.button('Show me the wines!'):
    with st.spinner('Should you have some wine now?'):
        
        time.sleep(2)
        # Instantiate selected wine traits
        if len(select_temptrait) == 0:
            selected_traits = all_traits
        else:
            selected_traits = select_temptrait

        # Run recommender model
        recommend_df = recommend_scores()
    
        # Instantiate traits filter
        trait_filter = ['title']

        # Add on any traits selected by user
        trait_filter.extend(selected_traits)

        # Create dataframe for wine name and traits
        df_temp_traits = df_wine_combi.drop(columns=['taster_name', 'points', 'variety', 'designation', 'winery', 'country', 'province', 'region_1', 'region_2', 'price', 'description',
                                                     'desc_wd_count', 'traits'])

        # Code to start filtering out wines with either one of the selected traits
        df_temp_traits = df_temp_traits[trait_filter]
        df_temp_traits['sum'] = df_temp_traits.sum(axis=1, numeric_only=True)
        df_temp_traits = df_temp_traits[df_temp_traits['sum'] != 0]

        # Merge the selected wines traits with recommend scores
        df_selectrec_temp = df_temp_traits.merge(recommend_df, on='title', how='left')

        # Merge the selected wines with recommendations with df on details
        df_selectrec_detail = df_selectrec_temp.merge(df_wine_combi, on='title', how='left')
        df_selectrec_detail.drop_duplicates(inplace=True)

        # Pull out the top 10 recommendations (raw)
        df_rec_raw = df_selectrec_detail.sort_values('est_match_pts', ascending=False).head(10)
        
        # Prepare the display for the top 10 recommendations
        df_rec_final = df_rec_raw[['title', 'country', 'province', 'variety', 'winery', 'points', 'price', 'traits', 'description']].reset_index(drop=True)
        df_rec_final.index = df_rec_final.index + 1
        df_rec_final['traits']=df_rec_final['traits'].str.replace(" ", " | ")
        df_rec_final.rename(columns={'title':'Name',
                                     'country':'Country',
                                     'province':'State/Province',
                                     'variety':'Type',
                                     'winery':'Winery',
                                     'points':'Rating',
                                     'price':'Price',
                                     'description':'Review',
                                     'traits':'Key Traits'}, inplace=True)
        st.balloons()
        st.dataframe(df_rec_final)