Machine Learning-Based Demand Forecasting for an FMCG Retailer

Supply chain management is an important part of the retail industry. It helps to control inventory levels, logistics expenses while providing adequate service to consumers, especially in large retail companies. There are many steps to managing the supply chain. One of the most important steps is having an accurate demand forecasting. Due to the wide variety of products in the retail sector and the many variables that affect sales, it can be difficult to forecast demand by hand. This study employs a machine learning-based approach to predict daily sales for each product-store tuple, determining store demand for a 15-day horizon. Predictions are based on statistical measures of past sales, product prices, promotions and time-dependent variables. Within the scope of this study, we create various features and use algorithms such as Linear Regression, Random Forest and LightGBM. We present the results of these models and compare our results with other approaches using various metrics.