A Comparison of ARIMA and Hybrid ARIMA-GRU Models for Forecasting Electric Bicycle Sales
bit-Tech · 10.08.2026 · 9 (1) · 599-611
DOI: 10.32877/bt.v9i1.3946 ↗Kurz gefasst
Fachartikel in der Zeitschrift bit-Tech, online erschienen am 10.08.2026. Verfasst von Herlambang Awan Irawan und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).
Abstract (Originalsprache)
This study aims to develop and evaluate a sales forecasting model for electric bicycles by comparing the performance of ARIMA model and a hybrid ARIMA-Gated Recurrent Unit (GRU) model. Accurate sales forecasting is essential for supporting inventory management, production planning, and strategic decision-making in the electric bicycle industry. The ARIMA model is employed to capture linear patterns and temporal dependencies in the time-series data, while the GRU model is utilized to learn nonlinear relationships that remain in the residual series generated by ARIMA. The research methodology begins with data preprocessing, including time-series decomposition to identify trend and seasonal components, stationarity testing using the Augmented Dickey–Fuller (ADF) test, differencing to achieve stationarity, and parameter identification through Autocorrelation-Function (ACF) and Partial Autocorrelation-Function (PACF) analyses. The optimal ARIMA model is selected based on the Akaike Information Criterion (AIC) and residual diagnostic tests to ensure model adequacy. The residual values obtained from the selected ARIMA model are then used as inputs for the GRU network. Hyperparameter optimization for the GRU model is conducted using a grid search approach to determine the most effective architecture and learning configuration. Experimental results indicate that the hybrid ARIMA–GRU model outperforms the standalone ARIMA model, achieving an MAE of 1.534962, an RMSE of 1.857339, and a MAPE of 4.992130%, compared with the ARIMA model, which produces an MAE of 2.643227, an RMSE of 3.244297, and a MAPE of 8.723934%. These findings demonstrate the effectiveness of combining statistical and deep learning approaches to improve sales forecasting accuracy.
Zitieren
APA
Irawan, H. A., Damaliana, A. T., & Syaifullah, W. (2026). A Comparison of ARIMA and Hybrid ARIMA-GRU Models for Forecasting Electric Bicycle Sales. bit-Tech, 9(1), 599-611. https://doi.org/10.32877/bt.v9i1.3946
Vancouver
Irawan HA, Damaliana AT, Syaifullah W. A Comparison of ARIMA and Hybrid ARIMA-GRU Models for Forecasting Electric Bicycle Sales. bit-Tech. 2026;9(1):599-611. doi:10.32877/bt.v9i1.3946
BibTeX
@article{Irawan2026,
author = {Irawan, Herlambang Awan and Damaliana, Aviolla Terza and Syaifullah, Wahyu},
title = {A Comparison of ARIMA and Hybrid ARIMA-GRU Models for Forecasting Electric Bicycle Sales},
journal = {bit-Tech},
year = {2026},
volume = {9},
number = {1},
pages = {599--611},
doi = {10.32877/bt.v9i1.3946}
}
EUVIMED