b.hmm_model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/Users/yibeihu/BestBuy_sales_forecast\r\n"
     ]
    }
   ],
   "source": [
    "!pwd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "import matplotlib\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import datetime\n",
    "\n",
    "sys.version\n",
    "sys.path.append(\"../../\")\n",
    "\n",
    "from src.hmm import sku_predict\n",
    "from src.utils import *\n",
    "import pickle\n",
    "import itertools\n",
    "\n",
    "from tqdm import tqdm\n",
    "from IPython.utils import io\n",
    "from joblib import Parallel, delayed\n",
    "\n",
    "pd.options.mode.chained_assignment = None"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Get all SKUs using HMM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('./data/final_sku_model_map.pkl', 'rb') as handle:\n",
    "    sku_period_dict1 = pickle.load(handle)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# extract SKU keys using HMM model to forecast\n",
    "hmm_keys = ['rmse_du_15', 'rmse_du_30', 'rmse_du_45', 'rmse_du_60', 'rmse_du_90', 'rmse_du_max']\n",
    "hmm_skus = [sku_period_dict1[sub_model] for sub_model in hmm_keys]\n",
    "hmm_skus = list(itertools.chain.from_iterable(hmm_skus))\n",
    "\n",
    "hmm_skus = list(set(hmm_skus))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prepare Data for training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load in data for training\n",
    "original_train = pd.read_csv(\"./data/train_validation_marker.csv\")\n",
    "original_train[\"SALES_DATE\"] = pd.to_datetime(original_train[\"SALES_DATE\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, '# SKU')"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Using a 90 day for HMM training.\n",
    "sku_in_scope = original_train[(original_train.Encoded_SKU_ID.isin(hmm_skus)) &(original_train.SALES_DATE>=\"2022-03-15\"\\\n",
    "                         )].groupby(\"Encoded_SKU_ID\")[\"DAILY_UNITS\"].nunique()\n",
    "\n",
    "sku_in_scope.hist()\n",
    "plt.title('Unique Sales Values Presented by Each SKU')\n",
    "plt.xlabel('# unique value')\n",
    "plt.ylabel('# SKU')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Test Using 1 SKU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Construct a dummy df validation set\n",
    "date_range = pd.date_range(start='08/01/2022', end='08/07/2022')\n",
    "dummy_forecast = pd.DataFrame(index=date_range)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Created 2 lag features\n",
      "Training on : ['DAILY_UNITS_lag1', 'DAILY_UNITS_lag2'] ['change_lag1', 'change_lag2']\n"
     ]
    }
   ],
   "source": [
    "#Specify the SKU and # of Gaussian Mixtures to construct model\n",
    "sku_id = 1\n",
    "n_comps = max(sku_in_scope[sku_id],10) #Limit to max 10 components\n",
    "\n",
    "#Initialize the model parameters\n",
    "sku_pred_model = sku_predict(original_train, sku_id)\n",
    "\n",
    "#train the model for this SKU\n",
    "feats_sku = sku_pred_model.get_features(n_lags=2)\n",
    "train1, _ = sku_pred_model.split_train_test(\"2022-07-31\")\n",
    "valid1 = dummy_forecast\n",
    "sku_pred_model.fit_hmm(train1, \"2022-03-01\", n_components1=n_comps)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Starting Prediction\n",
      "-----> Predicting for day:0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Even though the 'startprob_' attribute is set, it will be overwritten during initialization because 'init_params' contains 's'\n",
      "Even though the 'transmat_' attribute is set, it will be overwritten during initialization because 'init_params' contains 't'\n",
      "Fitting a model with 179 free scalar parameters with only 60 data points will result in a degenerate solution.\n",
      "Even though the 'means_' attribute is set, it will be overwritten during initialization because 'init_params' contains 'm'\n",
      "Even though the 'covars_' attribute is set, it will be overwritten during initialization because 'init_params' contains 'c'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-----> Predicting for day:1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Even though the 'startprob_' attribute is set, it will be overwritten during initialization because 'init_params' contains 's'\n",
      "Even though the 'transmat_' attribute is set, it will be overwritten during initialization because 'init_params' contains 't'\n",
      "Fitting a model with 179 free scalar parameters with only 60 data points will result in a degenerate solution.\n",
      "Even though the 'means_' attribute is set, it will be overwritten during initialization because 'init_params' contains 'm'\n",
      "Even though the 'covars_' attribute is set, it will be overwritten during initialization because 'init_params' contains 'c'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-----> Predicting for day:2\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Even though the 'startprob_' attribute is set, it will be overwritten during initialization because 'init_params' contains 's'\n",
      "Even though the 'transmat_' attribute is set, it will be overwritten during initialization because 'init_params' contains 't'\n",
      "Fitting a model with 179 free scalar parameters with only 60 data points will result in a degenerate solution.\n",
      "Even though the 'means_' attribute is set, it will be overwritten during initialization because 'init_params' contains 'm'\n",
      "Even though the 'covars_' attribute is set, it will be overwritten during initialization because 'init_params' contains 'c'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-----> Predicting for day:3\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Even though the 'startprob_' attribute is set, it will be overwritten during initialization because 'init_params' contains 's'\n",
      "Even though the 'transmat_' attribute is set, it will be overwritten during initialization because 'init_params' contains 't'\n",
      "Fitting a model with 179 free scalar parameters with only 60 data points will result in a degenerate solution.\n",
      "Even though the 'means_' attribute is set, it will be overwritten during initialization because 'init_params' contains 'm'\n",
      "Even though the 'covars_' attribute is set, it will be overwritten during initialization because 'init_params' contains 'c'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-----> Predicting for day:4\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Even though the 'startprob_' attribute is set, it will be overwritten during initialization because 'init_params' contains 's'\n",
      "Even though the 'transmat_' attribute is set, it will be overwritten during initialization because 'init_params' contains 't'\n",
      "Fitting a model with 179 free scalar parameters with only 60 data points will result in a degenerate solution.\n",
      "Even though the 'means_' attribute is set, it will be overwritten during initialization because 'init_params' contains 'm'\n",
      "Even though the 'covars_' attribute is set, it will be overwritten during initialization because 'init_params' contains 'c'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-----> Predicting for day:5\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Even though the 'startprob_' attribute is set, it will be overwritten during initialization because 'init_params' contains 's'\n",
      "Even though the 'transmat_' attribute is set, it will be overwritten during initialization because 'init_params' contains 't'\n",
      "Fitting a model with 179 free scalar parameters with only 60 data points will result in a degenerate solution.\n",
      "Even though the 'means_' attribute is set, it will be overwritten during initialization because 'init_params' contains 'm'\n",
      "Even though the 'covars_' attribute is set, it will be overwritten during initialization because 'init_params' contains 'c'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-----> Predicting for day:6\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>DAILY_UNITS_lag1</th>\n",
       "      <th>DAILY_UNITS_lag2</th>\n",
       "      <th>change_lag1</th>\n",
       "      <th>change_lag2</th>\n",
       "      <th>predicted1</th>\n",
       "      <th>predicted2</th>\n",
       "      <th>predicted</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2022-08-01</th>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2022-08-02</th>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>2.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2022-08-03</th>\n",
       "      <td>2.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2022-08-04</th>\n",
       "      <td>2.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>-2.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.55</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2022-08-05</th>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2022-08-06</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2022-08-07</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            DAILY_UNITS_lag1  DAILY_UNITS_lag2  change_lag1  change_lag2  \\\n",
       "2022-08-01               1.0               1.0          0.0          0.0   \n",
       "2022-08-02               0.0               1.0          2.0          1.0   \n",
       "2022-08-03               2.0               0.0         -1.0          1.0   \n",
       "2022-08-04               2.0               2.0         -1.0         -2.0   \n",
       "2022-08-05               0.0               1.0          0.0         -1.0   \n",
       "2022-08-06               0.0               0.0          0.0          0.0   \n",
       "2022-08-07               0.0               0.0          1.0          1.0   \n",
       "\n",
       "            predicted1  predicted2  predicted  \n",
       "2022-08-01         1.0         1.0       1.00  \n",
       "2022-08-02         2.0         2.0       2.00  \n",
       "2022-08-03         1.0         1.0       1.00  \n",
       "2022-08-04         1.0         0.0       0.55  \n",
       "2022-08-05         0.0         0.0       0.00  \n",
       "2022-08-06         0.0         0.0       0.00  \n",
       "2022-08-07         1.0         1.0       1.00  "
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sku_pred_model.predict(valid1, 15)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Parallel training for Batch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create an error df to store error values\n",
    "date_range = pd.date_range(start='08/01/2022', end='08/07/2022')\n",
    "error_prediction = pd.DataFrame(index=date_range)\n",
    "error_prediction[\"predicted\"] = np.nan\n",
    "# error_prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "def parallel_hmm_function(sku_id):\n",
    "    n_comps = max(sku_in_scope[sku_id],10) #Limit to max 10 components\n",
    "    sku_pred_model = sku_predict(original_train, sku_id)\n",
    "    with io.capture_output() as captured:\n",
    "        feats_sku = sku_pred_model.get_features(n_lags=2)\n",
    "        train1, _ = sku_pred_model.split_train_test(\"2022-07-31\")\n",
    "        valid1 = dummy_forecast\n",
    "        sku_pred_model.fit_hmm(train1, \"2022-03-01\", n_components1=n_comps)\n",
    "        try:\n",
    "            sku_pred_15 = sku_pred_model.predict(valid1, 15)\n",
    "        except:\n",
    "            sku_pred_15 = error_prediction\n",
    "            print(f\"Error for {sku_id} 15\")\n",
    "        try:\n",
    "            sku_pred_30 = sku_pred_model.predict(valid1, 30)\n",
    "        except:\n",
    "            sku_pred_30 = error_prediction\n",
    "            print(f\"Error for {sku_id} 30\")\n",
    "        try:\n",
    "            sku_pred_45 = sku_pred_model.predict(valid1, 45)\n",
    "        except:\n",
    "            sku_pred_45 = error_prediction\n",
    "            print(f\"Error for {sku_id} 45\")\n",
    "        try:\n",
    "            sku_pred_60 = sku_pred_model.predict(valid1, 60)\n",
    "        except:\n",
    "            sku_pred_60 = error_prediction\n",
    "            print(f\"Error for {sku_id} 60\")\n",
    "        try:\n",
    "            sku_pred_90 = sku_pred_model.predict(valid1, 90)\n",
    "        except:\n",
    "            sku_pred_90 = error_prediction\n",
    "            print(f\"Error for {sku_id} 90\")\n",
    "        try:\n",
    "            sku_pred_max = sku_pred_model.predict(valid1, sku_pred_model.X.shape[0])\n",
    "        except:\n",
    "            sku_pred_max = error_prediction\n",
    "            print(f\"Error for {sku_id} max\")\n",
    "        merge0 = sku_pred_15[[\"predicted\"]].merge(sku_pred_30[[\"predicted\"]], left_index=True, \\\n",
    "                                                  right_index=True, suffixes=(\"\", \"_30\"))\n",
    "        merge1 = merge0.merge(sku_pred_45[[\"predicted\"]], left_index=True, \\\n",
    "                              right_index=True, suffixes=(\"\", \"_45\"))\n",
    "        merge2 = merge1.merge(sku_pred_60[[\"predicted\"]], left_index=True, \\\n",
    "                              right_index=True, suffixes=(\"\", \"_60\"))\n",
    "        merge3 = merge2.merge(sku_pred_90[[\"predicted\"]], left_index=True, \\\n",
    "                              right_index=True, suffixes=(\"\", \"_90\"))\n",
    "        merge4 = merge3.merge(sku_pred_max[[\"predicted\"]], left_index=True, \\\n",
    "                              right_index=True, suffixes=(\"\", \"_max\"))\n",
    "        merge4[\"Encoded_SKU_ID\"] = [sku_id] * merge4.shape[0]\n",
    "        print(f\"prediction successful for {sku_id}\")\n",
    "        return(merge4,captured)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "  0%|          | 0/30 [00:00<?, ?it/s]\u001b[A\u001b[A\n",
      "\n",
      " 20%|██        | 6/30 [01:07<04:30, 11.26s/it]\u001b[A\u001b[A\n",
      "\n",
      " 30%|███       | 9/30 [02:38<05:57, 17.02s/it]\u001b[A\u001b[A\n",
      "\n",
      " 40%|████      | 12/30 [06:19<10:11, 33.99s/it]\u001b[A\u001b[A\n",
      "\n",
      " 50%|█████     | 15/30 [08:20<08:58, 35.88s/it]\u001b[A\u001b[A\n",
      "\n",
      " 60%|██████    | 18/30 [10:55<08:07, 40.65s/it]\u001b[A\u001b[A\n",
      "\n",
      " 70%|███████   | 21/30 [12:07<05:20, 35.63s/it]\u001b[A\u001b[A\n",
      "\n",
      " 80%|████████  | 24/30 [14:15<03:46, 37.70s/it]\u001b[A\u001b[A\n",
      "\n",
      " 90%|█████████ | 27/30 [15:01<01:32, 30.98s/it]\u001b[A\u001b[A\n",
      "\n",
      "100%|██████████| 30/30 [17:03<00:00, 34.12s/it]\u001b[A\u001b[A\n"
     ]
    }
   ],
   "source": [
    "all_preds = Parallel(n_jobs=3)(delayed(parallel_hmm_function)(sku_id) \\\n",
    "                               for sku_id in tqdm(sku_in_scope.index[:30]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "hmm_result = pd.concat([x[0] for x in all_preds])\n",
    "hmm_result.to_csv(f\"./data/hmm_result_final_0_30.csv\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}