What does this research mean for the field?

A stacking ensemble learning method that integrates Croston-based, machine learning, and deep learning models outperforms individual models in forecasting intermittent and lumpy demand for automotive spare parts. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

The aim is to enhance demand forecasting accuracy for automotive spare parts with irregular demand patterns.

April 3, 2026Open Access

An effective AI infused demand forecasting application for automotive spare parts industry: a real case from Turkey

Key Points

The aim is to enhance demand forecasting accuracy for automotive spare parts with irregular demand patterns.
Analysis of intermittent and lumpy demand structures using real data.
Comparison of Croston-based methods, machine learning, and deep learning models.
Integration of model outputs using stacking ensemble learning for improved forecasting accuracy.
The stacking ensemble method significantly outperformed traditional forecasting models.
Empirical results confirmed enhanced forecasting accuracy for intermittent and lumpy demand.
Statistical tests validate the effectiveness of the proposed innovative approach.

Abstract

The accuracy of demand forecasting in the automotive spare parts industry is critical to operational efficiency and financial performance. However, the irregular nature of spare parts demand makes forecasting processes quite complex. As traditional forecasting methods are unable to model this complex demand structure, researchers have developed more advanced and adaptive forecasting approaches. This study presents a comprehensive analysis to forecast the demand for products with intermittent and lumpy demand structure, which play a crucial role in the company’s sales process and operations, using real data from a company in the automotive spare parts industry. In addition, the results of Croston-based methods, ML, and DL models are compared, and an innovative forecasting approach is presented that integrates the outputs of these models with the stacking ensemble learning method. Empirical results and statistical tests confirm that the stacking method outperforms other models in forecasting intermittent and lumpy demand, highlighting the value of ensemble learning and advanced models.

Bookmark

View Full Paper