 |
International Society of Science and Applied Technologies |
| Demand Forecasting in the Aviation Aftermarket: A Gradient-Boosting Approach | ||||
| Author | Joffrey L. Leevy
|
|||
| Co-Author(s) | Chelsea M. Zuvieta; Taghi M. Khoshgoftaar
|
|||
| Abstract | This study investigates advanced gradient-boosting approaches to improve demand forecasting for used aircraft parts in the aviation aftermarket. Irregular demand patterns, high costs, and stringent service-level requirements make accurate forecasting a significant operational challenge. We utilized over 13,000 monthly aggregated data points and applied two gradient-boosting algorithms, CatBoost and XGBoost, to predict spare parts demand. Parts with more unstable demand patterns were excluded by applying Average Demand Interval (ADI) and Coefficient of Variation Squared (CV2) thresholds, and the data was augmented with engineered features. Our experimental results indicate that CatBoost, especially the CV2 filtering model, achieves the lowest Mean Absolute Error (MAE), outperforming XGBoost models. The proposed framework can integrate seamlessly into existing systems, providing a cost-effective, data-driven solution for improved supply-chain management in the aviation aftermarket.
|
|||
| Keywords | demand forecasting, aviation, CatBoost, XGBoost, feature selection, machine learning | |||
| Article #: RQD2025-183 | ||||
Proceedings of 30th ISSAT International Conference on Reliability & Quality in Design |