Business Analytics and Machine Learning-Based Demand Forecasting and Inventory Optimization in the Grocery Supply Chain

Abstract

This paper offers a detailed solution to the use of machine learning methods in demand forecasting and inventory optimization of the grocery supply chain. Random Forest (RF) and Extreme Gradient Boosting (XGBoost) models were used to estimate the demand of short-term products based on historical data of sales and inventory. Mean Absolute Error (MAE) was used to measure model performance to determine that RF delivered slightly more accurate forecasts than XGBoost. The projected demand was then used to optimise essential influent factors in inventory, such as reorder points and days of inventory, to be able to obtain sufficient in-store inventory, but with a minimal holding cost. The findings indicate that machine learning-based predictions may go a long way to improve inventory management, minimize stockouts, and make operations more efficient. These results provide practical suggestions to grocery stores that are interested in using the data-guided decision-making as a tool to improve their supply chain operations.