This project is collaborated by Chu-Yun Hsiao, Hao-Chun Niu, Shu-Yun Liu and Yu-Chun Peng.
We are data scientists who are passionate about the retail industry. We also notice that time series is a relatively important topic in sales prediction. Hence, we decided to take sales data from Walmart (2010~2013) as an example and created a time series model to predict 45 different Walmart stores’ sales. We ended up using the same parameters and machine learning methods to develop Forecast Models for each store. Thus, we have a total of 45 Forecasting Models for these 45 stores.
The visualization part of the project is done using Tableau dashboard. The access links are as below: https://public.tableau.com/app/profile/hao.chun.niu/viz/WalmartSalesForecastingPart1/WalmartSalesTimeSeriesEDAPart1 https://public.tableau.com/app/profile/hao.chun.niu/viz/WalmartSalesForecastingPart2/WalmartSalesTimeSeriesEDAPart2
Due to the large size of our data file, the Python codes and raw data files are stored on Google Drive. Provided the link as below: https://drive.google.com/drive/folders/1tmWvOgDHg5DUjxqgeoN_3eIpAKPJs9Hs?usp=sharing
- The Model file contains all the models attempted, including SARIMAX, Prophet, Prophet + XGBoost and Neural Network.
- The Raw Data file contains all the data files used. The original data source is from Kaggle. Link as below:
https://www.kaggle.com/divyajeetthakur/walmart-sales-prediction/.