Smart Retail Using Machine Learning for Demand Forecasting and Inventory Optimization

Considering the fact that the current level of complexity in the retail sphere caused by demand forecasting and inventory management, the given disable due to the outdated statistical methods such as moving averages and exponential smoothing lead to the increasing number of stockouts and overstock. The given technology fails to capture complex sales patterns and uses the limited scope of real-time data to manage the inventory. The scope of stockouts and overstock continues shaping the costs and decreasing customer satisfaction. In the regard, the given paper proposes the ML-based system which uses the most innovative algorithms such as Long Short-Term Memory and XGBoost to avoid such problems. The use of the specific data and algorithms helped to improve operational efficiency and decrease MAPE to 6.8% for such forecasts as the short-term ones. As a result, the system demonstrated improved inventory turnover ratio of 6.3 and the considerable decrease in the percentage of stockouts and overstock events. As a result of the investigation, it is possible to state that the given solution is effective since it helps to avoid forecasting errors and improve customer satisfaction.