Combining machine learning with special methods improves demand forecasts
Are Gradient Boosting Models Suitable for Intermittent Demand Forecasting?
Machine Learning
Summary
Predicting demand is hard for products that are rarely sold, leading to many days with zero sales. The authors studied if gradient boosting, a machine learning method, can help forecast these tricky cases. They found that specialized forecasting methods work best alone, but adding gradient boosting to them can improve accuracy by up to 10%. This means mixing smart machine learning with traditional methods can make better predictions for rare product sales.
What this means in practice
- •For inventory managers: Improve stock level predictions for rarely sold products by using combined machine learning and specialized forecasting models to reduce shortages and excesses.
- •For industrial planners: Enhance demand forecasts for intermittent items in supply chains by integrating gradient boosting with domain-specific methods to optimize ordering decisions.
Authors
Vladislav Kislinskii, Mazhar Hameed
Abstract
Demand forecasting is critical in modern industry, offering opportunities to reduce costs and gain competitive advantage through improved inventory management. However, forecasting becomes particularly challenging for products with intermittent demand, where demand occurs infrequently and time series contain many zero observations. Such dynamics are common across diverse sectors, such as industrial organizations, consumer goods, aviation, automotive, and electronics. Motivated by these challenges, this paper explores the potential of gradient boosting models to improve forecasting performance. We evaluate statistical, specialized, machine learning, and ensemble approaches across multiple datasets. The results show that specialized methods achieve the strongest performance among individual models, while gradient boosting on its own tends to underperform. However, combining a machine learning model with a specialized approach improves forecasting accuracy by up to 10%, demonstrating that even simple ensembles can outperform single models. Overall, the findings highlight the value of combining machine learning with domain-specific forecasting techniques for intermittent demand.