Comparative Analysis of Univariate and Multivariate Time Series Forecasting Models for Hard Token Inventory Planning in Banking

Authors

  • Risa Aprilia Universitas Bina Nusantara
  • Tuga Mauritsius Universitas Bina Nusantara

DOI:

https://doi.org/10.59261/jequi.v8i4.437

Keywords:

Demand Forecasting, Hard Token Inventory, Digital Banking, Univariate Forecasting, Multivariate Forecasting, LSTM, Temporal Fusion Transformer, CRISP-DM

Abstract

Background: Hard token availability is essential for secure transaction approval and service continuity in banking, yet quota-based stock planning may lead to stockouts, overstocking, and operational inefficiencies. This study compares univariate and multivariate time series forecasting models for branch-level hard token inventory planning.

Methods: This study aims to develop a forecasting model for branch-level hard Using the CRISP-DM framework, historical operational data from 197 branches were processed into a forecasting-ready dataset. ARIMA and Prophet were developed as univariate models, while Long Short-Term Memory (LSTM) and Temporal Fusion Transformer (TFT) incorporated operational, economic, and engineered temporal features. Model performance was evaluated using MAE, RMSE, MAPE, and R².

Results: TFT achieved the lowest MAE (5.997) and MAPE (11.56%), as well as the highest R² (0.7403), indicating the strongest overall performance across three of the four evaluation metrics. Prophet recorded the lowest RMSE (10.721), while ARIMA showed the weakest overall performance.

Conclusion: The results identify TFT as the preferred model for hard token demand forecasting, based on its lower absolute and percentage prediction errors and highest R² among the evaluated models. Although Prophet achieved a lower RMSE, TFT demonstrated stronger overall performance across the evaluation metrics. Integrating TFT-based forecasts into DigiMS can support more data-driven stock allocation, replenishment planning, and early identification of potential stock shortages and excess inventory at bank branches

Downloads

Download data is not yet available.

References

Acuña-Cid, H. A., Ahumada-Tello, E., Ovalle-Osuna, Ó. O., Evans, R., Hernández-Ríos, J. E., & Zambrano-Soto, M. A. (2025). CRISP-NET: Integration of the CRISP-DM model with network analysis. Machine Learning and Knowledge Extraction, 7(3), 1–38. https://doi.org/10.3390/make7030101

Bansal, A., Balaji, K., & Lalani, Z. (2025). Temporal encoding strategies for energy time series prediction. arXiv preprint arXiv:2503.15456.

Bhanja, S., & Das, A. (2018). Impact of data normalization on deep neural network for time series forecasting. arXiv preprint arXiv:1812.05519.

Bharath, A., Paduraru, A., & Gaidosch, T. (2024). Cyber resilience of the central bank digital currency ecosystem. International Monetary Fund.

Bueno, L. A., et al. (2024). Impacts of digitization on operational efficiency in the banking sector: Thematic analysis and research agenda proposal. International Journal of Information Management Data Insights, 4(1), 100230. https://doi.org/10.1016/j.jjimei.2024.100230

Doğan, A. (2025). Time series-based demand forecasting: A comparative analysis of Holt-Winters, SARIMA, and Prophet models on retail inventory data. İstanbul Ticaret Üniversitesi Sosyal Bilimler Dergisi, 24(53), 641–668. https://doi.org/10.46928/iticusbe.1725584

Goltsos, T. E., Syntetos, A. A., Glock, C. H., & Ioannou, G. (2022). Inventory–forecasting: Mind the gap. European Journal of Operational Research, 299(2), 397–419. https://doi.org/10.1016/j.ejor.2021.07.040

Hakizimana, S., Wairimu, M.-M. C., & Muathe, S. (2023). Digital banking transformation and performance-where do we stand. International Journal of Management Research and Emerging Sciences, 13(1), 179–196.

Kalaya, P., Termsuksawad, P., & Wasusri, T. (2023). Forecasting and inventory planning for irregular demand patterns. International Journal of Knowledge and Systems Science, 14(1).

Karmaker, S. (2025). Feature Engineering in Time Series Forecasting. Preprint, novembre.

Kiettikunwong, N., & Sangsarapun, W. (2025). G-Token implications and risks for the financial system under state-issued digital instruments in Thailand. Journal of Risk and Financial Management, 18(10), Article 555. https://doi.org/10.3390/jrfm18100555

Kim, M., Lee, J., Lee, C., & Jeong, J. (2022). Framework of 2D KDE and LSTM-based forecasting for cost-effective inventory management in smart manufacturing. Applied Sciences, 12(5), 2380. https://doi.org/10.3390/app12052380

Kontopoulou, V. I., Panagopoulos, A. D., Kakkos, I., & Matsopoulos, G. K. (2023). A review of ARIMA vs. machine learning approaches for time series forecasting in data driven networks. Future Internet, 15(8), Article 255. https://doi.org/10.3390/fi15080255

Lim, B., Arık, S. Ö., Loeff, N., & Pfister, T. (2021). Temporal fusion transformers for interpretable multi-horizon time series forecasting. International Journal of Forecasting, 37(4), 1748–1764. https://doi.org/10.1016/j.ijforecast.2021.03.012

Liu, R., & Vakharia, V. (2024). Optimizing supply chain management through BO-CNN-LSTM for demand forecasting and inventory management. Journal of Organizational and End User Computing, 36(1), 1–25. https://doi.org/10.4018/JOEUC.335591

Mauritsius, T. (2025). Extending the forecasting horizon of daily new COVID-19 cases using non-pharmaceutical measures and the effective reproduction number (Rt): A deep learning-based framework. Decision Science Letters, 14(2), 283–302. https://doi.org/10.5267/j.dsl.2025.1.008

Nguyen, T. N., Hossain, M. Z., Khan, M. M., Sharif, K. S., Das, R., & Haque, M. S. (2024). Product demand forecasting for inventory management with Freight Transportation Services Index using advanced neural networks algorithm. American Journal of Computing and Engineering, 7(4), 50–58. https://doi.org/10.47672/ajce.2432

Noor, N. M. M., Retnowardhani, A., Abd, M. L., & Saman, M. Y. M. (2013). Crime forecasting using ARIMA model and fuzzy alpha-cut. Journal of Applied Sciences, 13(1), 167–172. https://doi.org/10.3923/jas.2013.167.172

Schröer, C., Kruse, F., & Gómez, J. M. (2021). A systematic literature review on applying CRISP-DM process model. Procedia Computer Science, 181, 526–534. https://doi.org/10.1016/j.procs.2021.01.199

Wicaksono, B. A., & Mauritsius, T. (2025). Forecasting teaching materials using the autoregressive integrated moving average (ARIMA) method case study: Universitas Terbuka. International Journal of Engineering Trends and Technology, 73(6), 476–484. https://doi.org/10.14445/22315381/IJETT-V73I6P139

Downloads

Published

2026-10-22