Hybrid Ensemble sales forecasting model with Two-level adaptive correction based on the Kalman filter

Objective. The purpose of the study is to develop and experimentally validate a hybrid model for forecasting monthly sales volumes that is robust to nonstationarity, noise, and structural changes in time series characteristic of modern markets. Method. The paper proposes a combined forecasting model based on the integration of statistical methods (SARIMA, Holt–Winters) and the CatBoost machine learning algorithm with two-level adaptive correction using the Kalman filter. At the first level, the Kalman filter is applied for online calibration of individual forecasts of each model, which makes it possible to account for current changes in the dynamics of the series. At the second level, the Kalman filter is used to provide adaptive robustness of the weighted ensemble forecast. The model is implemented as a fully reproducible automated pipeline. Result. An experimental study was conducted using monthly sales data for the period 2020–2025 across 25 product items. The results showed that the proposed approach provides a reduction in forecasting errors by 6–12% according to MAE and MAPE metrics compared to baseline models, as well as a reduction in error variability by 3–5%, indicating increased forecast stability. The statistical significance of the improvements was confirmed by a paired t-test. Conclusion. The proposed hybrid model combines forecasting accuracy and robustness under conditions of an unstable future and can be effectively used to support managerial decision-making in planning and inventory management tasks.

Authors

Institutions

Publication Details

Journal
Herald of Dagestan State Technical University Technical Sciences
Published
2026-11-12
DOI
https://doi.org/10.21822/2073-6185-2026-53-3-103-111
Primary Topic
Forecasting Techniques and Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Hybrid Ensemble sales forecasting model with Two-level adaptive correction based on the Kalman filter

Andrey Polozov, Айбекова Элнура Айбековна, K. M. Khakimova
Herald of Dagestan State Technical University Technical Sciences
Forecasting Techniques and Applications
article

Hybrid Ensemble sales forecasting model with Two-level adaptive correction based on the Kalman filter

Andrey Polozov, Айбекова Элнура Айбековна, K. M. Khakimova
article en

Abstract

Objective. The purpose of the study is to develop and experimentally validate a hybrid model for forecasting monthly sales volumes that is robust to nonstationarity, noise, and structural changes in time series characteristic of modern markets. Method. The paper proposes a combined forecasting model based on the integration of statistical methods (SARIMA, Holt–Winters) and the CatBoost machine learning algorithm with two-level adaptive correction using the Kalman filter. At the first level, the Kalman filter is applied for online calibration of individual forecasts of each model, which makes it possible to account for current changes in the dynamics of the series. At the second level, the Kalman filter is used to provide adaptive robustness of the weighted ensemble forecast. The model is implemented as a fully reproducible automated pipeline. Result. An experimental study was conducted using monthly sales data for the period 2020–2025 across 25 product items. The results showed that the proposed approach provides a reduction in forecasting errors by 6–12% according to MAE and MAPE metrics compared to baseline models, as well as a reduction in error variability by 3–5%, indicating increased forecast stability. The statistical significance of the improvements was confirmed by a paired t-test. Conclusion. The proposed hybrid model combines forecasting accuracy and robustness under conditions of an unstable future and can be effectively used to support managerial decision-making in planning and inventory management tasks.

Herald of Dagestan State Technical University Technical SciencesVol. 53(3)
Ural Federal University (RU)
Forecasting Techniques and Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon