Prediction of Rossmann Store Sales

Introduction

This project is based on Kaggle Competition: Rossmann Store Sales. The goal is to predict sales of each Rossmann store in everyday of future three months. Based on the prediction, stores can arrange supply chain management, investment and workforce.

Rossmann provides sales and store data for 1,115 stores of nearly 3 years. The analysis process is as follows:

I firstly merge two tables into one. Because store conducting long-term promotion has different variables from that not conducting long-term promotion, I separately make prediction for these two types of stores. Besides, the test data does not have variable Number of Customers. However, this is a very important variable. Therefore, I predict number of customers for test dataset, and then predict sales. I use RMSPE to evaluate my prediction based on different algorithms and different time periods.

The following documents will give you more details about this project.

Documents

• Poster: Prediction of Rossmann Store Sales Poster.pdf
• Report: Prediction of Rossmann Store Sales Report.pdf
• Code
• Sample Code
• Preprocess: getdata.ipynb
• Predict Number of Customers: predictcustomers.ipynb
• Predict Sales: predictionwtcustomers.ipynb
• Download All: Prediction_of_Rossmann_Store_Sales_Code.zip

Thanks for the D3, Bootstrap and Glyphicon open source community.