Data-Driven Retail Excellence: Machine Learning for Demand Forecasting and Price Optimization

Abstract

Demand forecasting and price optimization are critical aspects of profitability for retailers in a supply chain. Retailers need to adopt innovative strategies to optimize pricing and increase profitability. This research paper proposes a price optimization approach for retailers using machine learning. The approach involves using linear regression to forecast demand incorporating price as an input, followed by price optimization taking into account inventory and perishability costs. The feasibility of using linear regression for price optimization for Stock Keeping Units (SKUs) is assessed using a feasibility index. The linear regression can predict the demand more accurately (23% Mean Absoulute Percentage Error (MAPE)) compared to exponential smoothing with optimised smoothing constant (47.09% MAPE) for 1000 SKUs. Also, the feasibility index can segregate the SKUs with an accuracy of 99%. The machine learning-based demand forecasting can assist retailers in accurately predicting customer demand and improving pricing decisions, while the feasibility index enables retailers to identify SKUs that require alternative pricing strategies.