Using Machine Learning to Predict Grocery Sales

Abstract

The purpose of this research is to construct a sales prediction model for grocery stores using linear regression. A forecasting model which can effectively predict the sales of a grocery store will help retailers stock the right quantity needed of their selling items and eventually compete better in the market. The present study uses four years’ worth of point-of-sale (POS) data from 54 grocery stores from Corporación Favorita, which is a large Ecuadorian-based grocery retailer to construct a sales prediction model that can predict the items demand for the upcoming year. XGboost (extreme gradient boosting) was used to build the prediction model. The constructed model can predict upcoming sales with an accuracy of 83%.