Quotidian Sales Forecasting using Machine Learning

Retailers have been experiencing a drop in their sales due to the rise of E-commerce facilities. This poses a problem where the retail stores need to efficiently manage and price their products to increase their sales. Hence the need for efficient sales prediction and dynamic pricing arises. A forecasting model which can effectively predict the sales of a retail store will help retailers compete in the market. With this intent, the paper proposes a model based on XGBoost whose learners are fitted to the store-product subsets with optimum parameters to increase the overall performance of sales prediction. The proposed model predicted sales for 10 stores with 50 products, with average MAPE, RMSE and R2 values of 11.98 %, 6.63 and 0.76 respectively. In addition, dynamic pricing is applied to the forecasted results which specifies the optimum price of a product based on its demand.