recsys_deploy_code.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "885d08d6-bc70-4a3d-b031-3ee58dc2f2e5",
   "metadata": {},
   "source": [
    "# Capstone Project: Wine Recommender\n",
    "---\n",
    "This notebook shows the code to create `requirements.txt` and `app.py` for Streamlit App deployment into Streamlit.<br>\n",
    "[Link to app](https://leewanxian-wine-recommender-app-44l3c7.streamlit.app/)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "9a415be1-bea9-4b5d-9384-d9bf8bad180d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Writing requirements.txt\n"
     ]
    }
   ],
   "source": [
    "%%writefile requirements.txt\n",
    "streamlit\n",
    "pandas\n",
    "numpy\n",
    "scikit-surprise"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "9adf7bf0-bef0-46e5-8438-488bcd0a4e4c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Writing app.py\n"
     ]
    }
   ],
   "source": [
    "%%writefile app.py\n",
    "# Import python lib\n",
    "import streamlit as st\n",
    "import time\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from surprise import Dataset, Reader\n",
    "from surprise import KNNBaseline\n",
    "\n",
    "# Import wine dataframes\n",
    "df_wine_model = pd.read_pickle('./data/df_wine_us_rate.pkl')\n",
    "df_wine_combi = pd.read_pickle('./data/df_wine_combi.pkl')\n",
    "\n",
    "# Instantiate the list of wine traits\n",
    "all_traits = ['almond', 'anise', 'apple', 'apricot', 'baked', 'baking_spices', 'berry', 'black_cherry', 'black_currant', 'black_pepper', 'black_tea', 'blackberry', 'blueberry', \n",
    "              'boysenberry', 'bramble', 'bright', 'butter', 'candy', 'caramel', 'cardamom', 'cassis', 'cedar', 'chalk', 'cherry', 'chocolate', 'cinnamon', 'citrus', 'clean', 'closed',\n",
    "              'clove', 'cocoa', 'coffee', 'cola', 'complex', 'concentrated', 'cranberry', 'cream', 'crisp', 'dark', 'dark_chocolate', 'dense', 'depth', 'dried_herb', 'dry', 'dust',\n",
    "              'earth', 'edgy', 'elderberry', 'elegant', 'fennel', 'firm', 'flower', 'forest_floor', 'french_oak', 'fresh', 'fruit', 'full_bodied', 'game', 'grapefruit', 'graphite',\n",
    "              'green', 'gripping', 'grippy', 'hearty', 'herb', 'honey', 'honeysuckle', 'jam', 'juicy', 'lavender', 'leafy', 'lean', 'leather', 'lemon', 'lemon_peel', 'length', 'licorice',\n",
    "              'light_bodied', 'lime', 'lush', 'meaty', 'medium_bodied', 'melon', 'milk_chocolate', 'minerality', 'mint', 'nutmeg', 'oak', 'olive', 'orange', 'orange_peel', 'peach',\n",
    "              'pear', 'pencil_lead', 'pepper', 'pine', 'pineapple', 'plum', 'plush', 'polished', 'pomegranate', 'powerful', 'purple', 'purple_flower', 'raspberry', 'refreshing',\n",
    "              'restrained', 'rich', 'ripe', 'robust', 'rose', 'round', 'sage', 'salt', 'savory', 'sharp', 'silky', 'smoke', 'smoked_meat', 'smooth', 'soft', 'sparkling', 'spice',\n",
    "              'steel', 'stone', 'strawberry', 'succulent', 'supple', 'sweet', 'tangy', 'tannin', 'tar', 'tart', 'tea', 'thick', 'thyme', 'tight', 'toast', 'tobacco', 'tropical_fruit',\n",
    "              'vanilla', 'velvety', 'vibrant', 'violet', 'warm', 'weight', 'wet_rocks', 'white', 'white_pepper', 'wood']\n",
    "\n",
    "#---------------------------------------------------------------------------------------------------------\n",
    "\n",
    "# Function to instantiate the model & return the est recsys scores\n",
    "def recommend_scores():\n",
    "    \n",
    "    # Instantiate reader & data for surprise\n",
    "    reader = Reader(rating_scale=(88, 100))\n",
    "    data = Dataset.load_from_df(df_wine_model, reader)\n",
    "    \n",
    "    # Instantiate recsys model\n",
    "    sim_options={'name':'cosine'}\n",
    "    model = KNNBaseline(k=35, min_k=1, sim_options=sim_options, verbose=False)\n",
    "\n",
    "    # Train & fit the data into model\n",
    "    train=data.build_full_trainset()\n",
    "    model.fit(train)\n",
    "\n",
    "    # Start the model to compute the best estimate match score on wine list\n",
    "    recommend_list = []\n",
    "    user_wines = df_wine_model[df_wine_model.taster_name == 'mockuser']['title'].unique()\n",
    "    not_user_wines = []\n",
    "    \n",
    "    for wine in df_wine_model['title'].unique():\n",
    "        if wine not in user_wines:\n",
    "            not_user_wines.append(wine)\n",
    "\n",
    "    for wine in not_user_wines:\n",
    "        wine_compatibility = []\n",
    "        prediction = model.predict(uid='mockuser', iid=wine)\n",
    "        wine_compatibility.append(prediction.iid)\n",
    "        wine_compatibility.append(prediction.est)\n",
    "        recommend_list.append(wine_compatibility)\n",
    "        \n",
    "    result_df = pd.DataFrame(recommend_list, columns = ['title', 'est_match_pts'])\n",
    "    \n",
    "    return result_df\n",
    "\n",
    "# Function for background image\n",
    "def add_bg_from_url():\n",
    "    st.markdown(\n",
    "        f\"\"\"\n",
    "        <style>\n",
    "        \n",
    "        [data-testid=\"stAppViewContainer\"] {{\n",
    "        background-image: url(\"https://images.pexels.com/photos/391213/pexels-photo-391213.jpeg\");\n",
    "        background-attachment: fixed;\n",
    "        background-size: cover      \n",
    "        }}\n",
    "        \n",
    "        [data-testid=\"stVerticalBlock\"] {{\n",
    "        background-color: rgba(255,255,255,0.5)\n",
    "        }}\n",
    "        \n",
    "        </style>\n",
    "        \"\"\",\n",
    "        unsafe_allow_html=True\n",
    "    )\n",
    "\n",
    "#----------------------------------------------------------------------------------------------------------\n",
    "\n",
    "st.title(\"Which wine should I get?\")\n",
    "st.write(\"By Lee Wan Xian\")\n",
    "st.write(\"[GitHub](https://github.com/leewanxian) | [LinkedIn](https://www.linkedin.com/in/wanxianlee)\")\n",
    "st.write(\"You can type the wine traits that you want in the dropdown list below\")\n",
    "add_bg_from_url()\n",
    "\n",
    "select_temptrait = st.multiselect(label = \" \", options = all_traits, label_visibility = \"collapsed\")\n",
    "\n",
    "if st.button('Show me the wines!'):\n",
    "    with st.spinner('Should you have some wine now?'):\n",
    "        \n",
    "        time.sleep(2)\n",
    "        # Instantiate selected wine traits\n",
    "        if len(select_temptrait) == 0:\n",
    "            selected_traits = all_traits\n",
    "        else:\n",
    "            selected_traits = select_temptrait\n",
    "\n",
    "        # Run recommender model\n",
    "        recommend_df = recommend_scores()\n",
    "    \n",
    "        # Instantiate traits filter\n",
    "        trait_filter = ['title']\n",
    "\n",
    "        # Add on any traits selected by user\n",
    "        trait_filter.extend(selected_traits)\n",
    "\n",
    "        # Create dataframe for wine name and traits\n",
    "        df_temp_traits = df_wine_combi.drop(columns=['taster_name', 'points', 'variety', 'designation', 'winery', 'country', 'province', 'region_1', 'region_2', 'price', 'description',\n",
    "                                                     'desc_wd_count', 'traits'])\n",
    "\n",
    "        # Code to start filtering out wines with either one of the selected traits\n",
    "        df_temp_traits = df_temp_traits[trait_filter]\n",
    "        df_temp_traits['sum'] = df_temp_traits.sum(axis=1, numeric_only=True)\n",
    "        df_temp_traits = df_temp_traits[df_temp_traits['sum'] != 0]\n",
    "\n",
    "        # Merge the selected wines traits with recommend scores\n",
    "        df_selectrec_temp = df_temp_traits.merge(recommend_df, on='title', how='left')\n",
    "\n",
    "        # Merge the selected wines with recommendations with df on details\n",
    "        df_selectrec_detail = df_selectrec_temp.merge(df_wine_combi, on='title', how='left')\n",
    "        df_selectrec_detail.drop_duplicates(inplace=True)\n",
    "\n",
    "        # Pull out the top 10 recommendations (raw)\n",
    "        df_rec_raw = df_selectrec_detail.sort_values('est_match_pts', ascending=False).head(10)\n",
    "        \n",
    "        # Prepare the display for the top 10 recommendations\n",
    "        df_rec_final = df_rec_raw[['title', 'points', 'price', 'variety', 'country', 'province', 'winery', 'description', 'traits']].reset_index(drop=True)\n",
    "        df_rec_final.index = df_rec_final.index + 1\n",
    "        df_rec_final['traits']=df_rec_final['traits'].str.replace(\" \", \" | \")\n",
    "        df_rec_final.rename(columns={'title':'Name',\n",
    "                                     'country':'Country',\n",
    "                                     'province':'State/Province',\n",
    "                                     'variety':'Type',\n",
    "                                     'winery':'Winery',\n",
    "                                     'points':'Rating (Out of 100)',\n",
    "                                     'price':'Price',\n",
    "                                     'description':'Review',\n",
    "                                     'traits':'Key Traits'}, inplace=True)\n",
    "        st.balloons()\n",
    "        st.dataframe(df_rec_final.style.format({\"Price\": \"${:,.2f}\"}))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:capstone]",
   "language": "python",
   "name": "conda-env-capstone-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}