walmart-sales-forecasting-pycaret-library.ipynb

{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.7.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"<p style='text-align:center;\n          font-family: sans-serif;\n          font-weight:bold;\n          color:black;\n          font-size:40px;\n          margin: 30px;'> Walmart Sales \n    <font color='#128EB3'; size='7'>Forecasting</font></p>\n<p style=\"text-align:center;\n          font-family: sans-serif;\n          font-weight:bold;\n          color:#616161;\n          font-size:20px;\n          margin: 30px;\">PyCaret library 📈</p>","metadata":{}},{"cell_type":"markdown","source":"<h3 style=\"text-align: left; background-color: #128EB3; font-family:Trebuchet MS; color: white; padding: 14px; line-height: 1; border-radius:10px\"> Notebook goals🔎</h3>","metadata":{}},{"cell_type":"markdown","source":"In this first version of the project the objective will be: \n\n* Predicting store-level sales behavior through time series modeling with some of the libraries available in Python\n\n**Note: I have divided the project into several notebooks for each of the models used for sales forecasting. I leave the link below:**\n\n\n* [Exploratory Data Analysis](https://www.kaggle.com/code/armandodelahoya/walmart-sales-forecasting-eda/notebook)\n* [Prophet model](https://www.kaggle.com/code/armandodelahoya/walmart-sales-forecasting-prophet-model/notebook)\n* [Neural prophet model](https://www.kaggle.com/code/armandodelahoya/walmart-sales-forecasting-neural-model/notebook)\n* [PyCaret library](https://www.kaggle.com/code/armandodelahoya/walmart-sales-forecasting-pycaret-library/notebook)\n* [Power BI Dashboard](https://app.powerbi.com/view?r=eyJrIjoiNWFjYWU0ZTktNjFiYS00Zjg3LTg0M2MtOTMzM2M0MjQ5OThmIiwidCI6IjAyNDlhNTcxLWI5YTItNGNhMi1iOTNiLTIwYzc3MDg4ZjA4YiJ9)\n* [Github repository](https://github.com/ArmandoLazalde/Walmart-Sales-Forecasting)","metadata":{}},{"cell_type":"markdown","source":"<a id=\"top\"></a>\n<div class=\"list-group\" id=\"list-tab\" role=\"tablist\">\n    <h3 style=\"text-align: left; background-color: #128EB3; font-family:Trebuchet MS; color: white; padding: 14px; line-height: 1; border-radius:10px\"><b>Table of Contents</b></h3>\n    \n* [1. Import necessary libraries](#1)\n* [2. Best store analysis](#2)\n* [3. All stores analysis](#3)","metadata":{}},{"cell_type":"markdown","source":"<a id=\"1\"></a>\n<h3 style=\"text-align: left;\n           background-color: #128EB3;\n           font-family:Trebuchet MS; \n           color: white; \n           padding: 14px; \n           line-height: 1; \n           border-radius:10px\"> 1. Import necessary libraries\n</h3>","metadata":{}},{"cell_type":"code","source":"pip install pycaret-ts-alpha","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:16:34.589437Z","iopub.execute_input":"2023-03-11T06:16:34.589973Z","iopub.status.idle":"2023-03-11T06:17:06.393891Z","shell.execute_reply.started":"2023-03-11T06:16:34.589907Z","shell.execute_reply":"2023-03-11T06:17:06.392417Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import numpy as np\nimport pandas as pd\nfrom datetime import datetime\nfrom zipfile import ZipFile\n\n#pycaret\ndef what_is_installed():\n    from pycaret import show_versions\n    show_versions()\n\ntry:\n    what_is_installed()\nexcept ModuleNotFoundError:\n    !pip install pycaret\n    what_is_installed()\n    \nimport time\nfrom pycaret.time_series import TSForecastingExperiment\nfrom pycaret.time_series import *\n\n#Metrics\nfrom sklearn.metrics import mean_absolute_error, mean_absolute_percentage_error, mean_squared_error, r2_score","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:06.398745Z","iopub.execute_input":"2023-03-11T06:17:06.399210Z","iopub.status.idle":"2023-03-11T06:17:13.044409Z","shell.execute_reply.started":"2023-03-11T06:17:06.399157Z","shell.execute_reply":"2023-03-11T06:17:13.043057Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"with ZipFile('/kaggle/input/walmart-recruiting-store-sales-forecasting/features.csv.zip') as f:\n    f.extractall(path='walmart-recruiting-store-sales-forecasting')\n    \nwith ZipFile('/kaggle/input/walmart-recruiting-store-sales-forecasting/sampleSubmission.csv.zip') as f:\n    f.extractall(path='walmart-recruiting-store-sales-forecasting') \n    \nwith ZipFile('/kaggle/input/walmart-recruiting-store-sales-forecasting/test.csv.zip') as f:\n    f.extractall(path='walmart-recruiting-store-sales-forecasting') \n    \nwith ZipFile('/kaggle/input/walmart-recruiting-store-sales-forecasting/train.csv.zip') as f:\n    f.extractall(path='walmart-recruiting-store-sales-forecasting') ","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:13.045944Z","iopub.execute_input":"2023-03-11T06:17:13.046350Z","iopub.status.idle":"2023-03-11T06:17:13.244554Z","shell.execute_reply.started":"2023-03-11T06:17:13.046312Z","shell.execute_reply":"2023-03-11T06:17:13.243378Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_df = pd.read_csv('/kaggle/working/walmart-recruiting-store-sales-forecasting/train.csv')\nfeatures_df = pd.read_csv('/kaggle/working/walmart-recruiting-store-sales-forecasting/features.csv') #index_col='Date')\nstores_df = pd.read_csv('/kaggle/input/walmart-recruiting-store-sales-forecasting/stores.csv')\ntest_df = pd.read_csv('/kaggle/working/walmart-recruiting-store-sales-forecasting/test.csv')","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:13.247318Z","iopub.execute_input":"2023-03-11T06:17:13.247666Z","iopub.status.idle":"2023-03-11T06:17:14.390408Z","shell.execute_reply.started":"2023-03-11T06:17:13.247633Z","shell.execute_reply":"2023-03-11T06:17:14.389037Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"data = train_df.merge(features_df, on=['Store', 'Date', 'IsHoliday'], how = 'inner')\nfinal_df = data.merge(stores_df, on=['Store'], how='inner')\nfinal_df.head(5)","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:14.392124Z","iopub.execute_input":"2023-03-11T06:17:14.392495Z","iopub.status.idle":"2023-03-11T06:17:14.695843Z","shell.execute_reply.started":"2023-03-11T06:17:14.392459Z","shell.execute_reply":"2023-03-11T06:17:14.694979Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"final_df[\"Date\"] = pd.to_datetime(final_df[\"Date\"], dayfirst=True)","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:14.697074Z","iopub.execute_input":"2023-03-11T06:17:14.697875Z","iopub.status.idle":"2023-03-11T06:17:14.768534Z","shell.execute_reply.started":"2023-03-11T06:17:14.697838Z","shell.execute_reply":"2023-03-11T06:17:14.767092Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"<a id=\"1\"></a>\n<h3 style=\"text-align: left;\n           background-color: #128EB3;\n           font-family:Trebuchet MS; \n           color: white; \n           padding: 14px; \n           line-height: 1; \n           border-radius:10px\"> 2. Best store analysis\n</h3>","metadata":{}},{"cell_type":"code","source":"#Store 20 analysis\nstore20 = final_df[final_df['Store'] == 20]\nstore20 = store20[['Date','Weekly_Sales']]\nstore20 = store20.groupby('Date').sum()\nstore20","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:14.770553Z","iopub.execute_input":"2023-03-11T06:17:14.771004Z","iopub.status.idle":"2023-03-11T06:17:14.864435Z","shell.execute_reply.started":"2023-03-11T06:17:14.770931Z","shell.execute_reply":"2023-03-11T06:17:14.863154Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# We want to forecast the next 23 weeks of data and we will use 3 fold cross-validation to test the models.\nfh = 23\nfold = 3\n\n# Global Plot Settings\nfig_kwargs={'renderer': 'notebook'}","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:14.866061Z","iopub.execute_input":"2023-03-11T06:17:14.866678Z","iopub.status.idle":"2023-03-11T06:17:14.872523Z","shell.execute_reply.started":"2023-03-11T06:17:14.866638Z","shell.execute_reply":"2023-03-11T06:17:14.871034Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"eda = TSForecastingExperiment()\neda.setup(store20, fh=fh, fig_kwargs=fig_kwargs)","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:14.874318Z","iopub.execute_input":"2023-03-11T06:17:14.875300Z","iopub.status.idle":"2023-03-11T06:17:15.481531Z","shell.execute_reply.started":"2023-03-11T06:17:14.875243Z","shell.execute_reply":"2023-03-11T06:17:15.480435Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"eda.plot_model()","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:15.485489Z","iopub.execute_input":"2023-03-11T06:17:15.486561Z","iopub.status.idle":"2023-03-11T06:17:17.321694Z","shell.execute_reply.started":"2023-03-11T06:17:15.486520Z","shell.execute_reply":"2023-03-11T06:17:17.320547Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Classical decomposition\neda.plot_model(plot=\"decomp\", data_kwargs={'seasonal_period': 12})","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:17.323365Z","iopub.execute_input":"2023-03-11T06:17:17.323716Z","iopub.status.idle":"2023-03-11T06:17:17.525261Z","shell.execute_reply.started":"2023-03-11T06:17:17.323682Z","shell.execute_reply":"2023-03-11T06:17:17.524051Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Show the train-test splits on the dataset\n# Internally split - len(fh) as test set, remaining used as test set\neda.plot_model(plot=\"train_test_split\", fig_kwargs={\"height\": 400, \"width\": 900})\n\n# Show the Cross Validation splits inside the train set\n# The blue dots represent the training data for each fold.\n# The orange dots represent the validation data for each fold\neda.plot_model(plot=\"cv\", fig_kwargs={\"height\": 400, \"width\": 900})","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:17.526825Z","iopub.execute_input":"2023-03-11T06:17:17.527227Z","iopub.status.idle":"2023-03-11T06:17:18.845146Z","shell.execute_reply.started":"2023-03-11T06:17:17.527186Z","shell.execute_reply":"2023-03-11T06:17:18.843856Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"exp = TSForecastingExperiment()\nexp.setup(store20, fh=fh, fig_kwargs=fig_kwargs) ","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:18:29.412218Z","iopub.execute_input":"2023-03-11T06:18:29.412699Z","iopub.status.idle":"2023-03-11T06:18:29.511173Z","shell.execute_reply.started":"2023-03-11T06:18:29.412645Z","shell.execute_reply":"2023-03-11T06:18:29.509994Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#### Initial Training ----\nmodel = exp.create_model(\"prophet\", freq='W-Fri')\nexp.plot_model(model)","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:18:33.555789Z","iopub.execute_input":"2023-03-11T06:18:33.557440Z","iopub.status.idle":"2023-03-11T06:18:43.766391Z","shell.execute_reply.started":"2023-03-11T06:18:33.557362Z","shell.execute_reply":"2023-03-11T06:18:43.765392Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"exp.predict_model(model)","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:18:56.956094Z","iopub.execute_input":"2023-03-11T06:18:56.957151Z","iopub.status.idle":"2023-03-11T06:18:58.159198Z","shell.execute_reply.started":"2023-03-11T06:18:56.957094Z","shell.execute_reply":"2023-03-11T06:18:58.157999Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#Final model\nfinal_model = exp.finalize_model(model)\nexp.plot_model(final_model)\nexp.predict_model(final_model)","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:19:12.907390Z","iopub.execute_input":"2023-03-11T06:19:12.907843Z","iopub.status.idle":"2023-03-11T06:19:22.657207Z","shell.execute_reply.started":"2023-03-11T06:19:12.907799Z","shell.execute_reply":"2023-03-11T06:19:22.656058Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"_ = exp.save_model(final_model, \"my_final_model\")","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:19:38.197744Z","iopub.execute_input":"2023-03-11T06:19:38.198255Z","iopub.status.idle":"2023-03-11T06:19:38.494590Z","shell.execute_reply.started":"2023-03-11T06:19:38.198201Z","shell.execute_reply":"2023-03-11T06:19:38.493357Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"exp_load = TSForecastingExperiment()\nloaded_model = exp_load.load_model(\"my_final_model\")","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:19:46.151638Z","iopub.execute_input":"2023-03-11T06:19:46.152164Z","iopub.status.idle":"2023-03-11T06:19:46.169402Z","shell.execute_reply.started":"2023-03-11T06:19:46.152108Z","shell.execute_reply":"2023-03-11T06:19:46.168000Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Should match predictions from before the save and load\nexp_load.predict_model(loaded_model)","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:19:48.724295Z","iopub.execute_input":"2023-03-11T06:19:48.725393Z","iopub.status.idle":"2023-03-11T06:19:49.889034Z","shell.execute_reply.started":"2023-03-11T06:19:48.725343Z","shell.execute_reply":"2023-03-11T06:19:49.887728Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Example here shows forecasting out 36 weeks \nexp.plot_model(estimator=final_model, data_kwargs={'fh': 36}) ","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:19:53.815722Z","iopub.execute_input":"2023-03-11T06:19:53.816224Z","iopub.status.idle":"2023-03-11T06:19:55.098451Z","shell.execute_reply.started":"2023-03-11T06:19:53.816171Z","shell.execute_reply":"2023-03-11T06:19:55.097160Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"#exp.models()\nexp.compare_models()","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:19.215125Z","iopub.status.idle":"2023-03-11T06:17:19.215672Z","shell.execute_reply.started":"2023-03-11T06:17:19.215381Z","shell.execute_reply":"2023-03-11T06:17:19.215411Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"<a id=\"3.\"></a>\n<h3 style=\"text-align: left; background-color: #128EB3; font-family:Trebuchet MS; color: white; padding: 14px; line-height: 1; border-radius:6px\"> 3. Analysis for all stores</h3>","metadata":{}},{"cell_type":"code","source":"results = {}\n\nfor store in final_df['Store'].unique():\n    \n    #Analysis per store\n    store_x = final_df[final_df['Store'] == store]\n    store_x = store_x[['Date','Weekly_Sales']]\n    store_x = store_x.groupby('Date').sum()\n    \n    fh = 23\n    fold = 3\n\n    exp = TSForecastingExperiment()\n    exp.setup(store_x, fh=fh)\n    model = exp.create_model(\"prophet\", freq='W-Fri')\n    \n    ypred = exp.predict_model(model)\n    \n    y_true = store_x['Weekly_Sales'].tail(23).values\n    y_pred = ypred['y_pred'].values\n    \n    mape = (mean_absolute_percentage_error(y_true, y_pred).round(4))*100\n    mae = mean_absolute_error(y_true, y_pred).round(4)\n    \n    results[store] = [mape, mae]","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:19.217154Z","iopub.status.idle":"2023-03-11T06:17:19.217712Z","shell.execute_reply.started":"2023-03-11T06:17:19.217415Z","shell.execute_reply":"2023-03-11T06:17:19.217446Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"results = pd.DataFrame(results)\nresults = results.transpose()\nresults.reset_index(inplace=True)\nresults.columns = ['Store', 'MAPE', 'MAE']\nresults","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:19.219342Z","iopub.status.idle":"2023-03-11T06:17:19.221460Z","shell.execute_reply.started":"2023-03-11T06:17:19.221249Z","shell.execute_reply":"2023-03-11T06:17:19.221273Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"results.to_csv(\"pycaret_results.csv\", index=False)","metadata":{"execution":{"iopub.status.busy":"2023-03-11T06:17:19.222546Z","iopub.status.idle":"2023-03-11T06:17:19.223138Z","shell.execute_reply.started":"2023-03-11T06:17:19.222928Z","shell.execute_reply":"2023-03-11T06:17:19.222949Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]}]}