A Fair Comparative Framework for Time-Series Forecasting Using ARIMAX, XGBoost, and LSTM: Evidence from Libya
DOI:
https://doi.org/10.65405/y1kqxv30Keywords:
Experimental Study, Monetary Base Forecasting, ARIMAX, XGBoost, LSTM, Diebold–Mariano Test, Libya, Fair Comparison, Sensitivity AnalysisAbstract
This experimental study provides a comparative analysis of three forecasting approaches ARIMAX (econometric), XGBoost (machine learning), and LSTM (deep learning) for predicting the Libyan monetary base using actual monthly data from January 2004 to April 2026 (268 observations). All forecasts are based on actual historical data with no simulated or future values. Accurate forecasting of the Libyan monetary base is critical for economic stability in a fragile, oil‑dependent economy where forecast errors can have significant financial repercussions. To ensure fairness, all models receive identical multivariate input features (net foreign assets, net domestic assets, and lagged monetary base values). Previous studies have focused primarily on stable economies, leaving a gap in understanding forecasting methods applicable to volatile contexts such as Libya. The experiment follows a rigorous framework with a training period (2004–2020) and a testing period (2021–2026). Performance is evaluated using MAE, RMSE, and MAPE, and statistical significance is assessed using the Diebold–Mariano test. Results show that ARIMAX(2,1,2) substantially outperforms both XGBoost and LSTM, achieving a MAPE of 25.36% compared to 28.41% (XGBoost) and 49.46% (LSTM). The Diebold–Mariano test confirms statistically significant differences (ARIMAX vs XGBoost: DM = -5.97, p < 0.001; ARIMAX vs LSTM: DM = -13.47, p < 0.001). Feature importance indicates that lagged monetary base values dominate predictions, while LSTM struggles with structural breaks and limited sample size. Sensitivity analysis shows that a 12‑month window is optimal for both LSTM and XGBoost. This study contributes an unbiased, reproducible experimental benchmark for monetary forecasting in fragile, oil‑dependent economies.
Downloads
References
Abbasimehr, H., Behboodi, A., & Bahrini, A. (2024). A novel hybrid model to forecast seasonal and chaotic time series. Expert Systems with Applications, 238, 122461.
Agaal, A., Essgaer, M., Farkash, H. M., & Othman, Z. A. (2025). Data-driven Insights for Informed Decision-Making: Applying LSTM Networks for Robust Electricity Forecasting in Libya. International Journal of Intelligent Systems and Applications (IJISA), 17(3), 65–89.
Al Feki, E., & Neji, J. (2024). Statistical modelling to assessing and enhancing road traffic safety in Tripoli, Libya: A systematic approach. Journal of Engineering Research, 12(4), 659-669.
Alrawayati, H., & Tökeşer, Ü. (2021). PARKINSON’S DISEASE DIAGNOSIS BASED ON THE CONVOLUTIONAL NEURAL NETWORK AND PARTICLE SWARM OPTIMIZATION ALGORITHM. Asian Journal of Mathematics and Computer Research, 28(1), 26-37.
Al‑Saffar, M., & Al‑Shammari, M. (2022). Structural breaks and forecasting performance in resource-rich countries. Resources Policy, 76, 102567.
Alssager, M., & Othman, Z. A. (2016). Taguchi-based parameter setting of cuckoo search algorithm for capacitated vehicle routing problem. In Advances in Machine Learning and Signal Processing: Proceedings of MALSIP 2015 (pp. 71-79). Cham: Springer International Publishing.
Ben Dalla, L., Medeni, T. M., Agila, A. A., & Medeni, İ. M. (2024). Architectural Synergy: Investigating the Role of Artificial Neural Networks in Enabling Deep Learning. The International Journal of Engineering & Information Technology (IJEIT), 12(1), 96-103.
Box, G. E. P., & Jenkins, G. M. (1976). Time Series Analysis: Forecasting and Control. Holden‑Day.
Chantar, H., Tubishat, M., Essgaer, M., & Mirjalili, S. (2021). Hybrid binary dragonfly algorithm with simulated annealing for feature selection. SN computer science, 2(4), 295.
Chen, T., & Guestrin, C. (2016). XGBoost: A scalable tree boosting system. Proceedings of the 22nd ACM SIGKDD, 785–794.
Coroneo, L., & Iacone, F. (2024). Testing for equal predictive accuracy with strong dependence. International Journal of Forecasting. (Accepted/In Press).
Dalla, L. O. B., Essgaer, M., Jetlawei, S. S., EL-sseid, M., Alsharif, A., & Agila, A. A. A. (2026). Local Precision and Global Harmony: A Comparative Literature Review (LR) Framework for Taylor and Fourier Series in Engineering Modeling. Al-Farooq Journal of Sciences, 2(1), 275-304.
Dalla, L. O. B., Essgaer, M., Jetlawei, S. S., EL-sseid, M., Alsharif, A., & Agila, A. A. A. (2026). Local Precision and Global Harmony: A Comparative Literature Review (LR) Framework for Taylor and Fourier Series in Engineering Modeling. Al-Farooq Journal of Sciences, 2(1), 275-304.
Dalla, L. O. B., Karal, Ö., & Degirmenciyi, A. (2025). Leveraging LSTM for adaptive intrusion detection in IoT networks: a case study on the RT-IoT2022 dataset implemented on CPU computer device machine. In 5th International Conference on Engineering, Natural and Social Sciences April (pp. 15-16).
Diebold, F. X., & Mariano, R. S. (1995). Comparing predictive accuracy. Journal of Business & Economic Statistics, 13(3), 253–263.
El-Sayed, M., & Hassan, T. (2023). Forecasting monetary aggregates in conflict-affected states: A case study of Libya and Syria. Middle East Development Journal, 15(1), 88–112.
Elsherif, M. (2024). Modelling Inflation Dynamics and Global Oil Price Shocks in OAPEC Countries: TVP-VAR. International Journal of Energy Economics and Policy, 14(3), 51–69.
Gonzalez, C., & Fernandez, R. (2024). XGBoost feature importance in financial time series: Interpretation and limitations. IEEE Transactions on Neural Networks and Learning Systems, 35(3), 2890–2902.
Tareq Alnnale. (2026). From Reactive to Proactive Governance: A Hybrid LSTM–Gradient Boosting Architecture for Real-Time Anomaly Signal Detection in Multi-Store Retail Supply Chain Decision Systems. Al-Farooq Journal of Sciences, 2(1), 987-1005.
Harvey, D. I., Leybourne, S. J., & Zu, Y. (2025). Testing for equal average forecast accuracy in possibly unstable environments. Journal of Business & Economic Statistics, 43(3), 643–656.
Hawa Ahmed Alrawayati, Ümit Tokeşer. (2025).Spectral Integral Variation of Graph Theory. Asian Journal of Mathematics and Computer Research.32, Issue, 2. Pages(151-160). https://www.elibrary.ru/item.asp?id=82163806
Hawa Ahmed Alrawayati, Ümit Tokeşer. (2025).Spectral Integral Variation of Graph Theory. Asian Journal of Mathematics and Computer Research.32, Issue, 2. Pages(151-160). https://www.elibrary.ru/item.asp?id=82163806.
Hawa Alrawayati (2020). Development of High Efficiency Optimization Algorithm based on New Topology in Particle Swarm Optimization for Parkinson’s Disease. IOSR Journal of Mathematics (IOSR-JM). 8
Hawa Alrawayati. (2013). (المؤثرات الخطية المحدودة على فضاء هلبرت) • Finite linear operators on Hilbert spaces.مجلة جامعة الزيتونة.184-193
Hawa Alrawayati. (2016). (المعادلة التكاملية ونواة المؤثر ) Integral Equation and Kernel Operator.مجلة الساتل - جامعة مصراته. 63 – 76
Hawa Alrawayati. (2016). Integral Equation and Kernel Operator. Al-Satel Journal - Misrata University. 63-76
Henderi, H., & Sofiana, S. (2025). Comparative Study of Traditional and Modern Models in Time Series Forecasting for Inflation Prediction. International Journal of Applied Informatics and Management, 5(3), 155–167.
Hochreiter, S., & Schmidhuber, J. (1997). Long short‑term memory. Neural Computation, 9(8), 1735–1780.
Jeguirim, K., & Ben Salem, L. (2024). Unveiling extreme dependencies between oil price shocks and inflation in Tunisia: Insights from a copula dcc garch approach. MPRA Paper No. 121616. University Library of Munich.
Karal, Ö. (2020). Performance comparison of different kernel functions in SVM for different k value in k-fold cross-validation. In 2020 innovations in intelligent systems and applications conference (ASYU) (pp. 1-5). IEEE.
Karal, Ö. (2020). Performance comparison of different kernel functions in SVM for different k value in k-fold cross-validation. In 2020 innovations in intelligent systems and applications conference (ASYU) (pp. 1-5). IEEE.
Karal, Ö. (2024). Comparative performance analysis of epsilon-insensitive and pruningbased algorithms for sparse least squares support vector regression. Sigma Journal of Engineering and Natural Sciences, 42(2), 578-589.
Khanh, M. Q., & Tuan, T. A. (2024). ARIMA, XGBoost, LSTM — Which One is Better for Forecasting?. Journal of Advanced Research in Applied Sciences and Engineering Technology, 43(1), 101-115.
Kumar, S., & Haider, A. (2022). LSTM vs ARIMA: A comparative study of forecasting models for macroeconomic indicators in developing economies. International Journal of Forecasting, 38(4), 1421–1438.
Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2018). Statistical and machine learning forecasting methods: Concerns and ways forward. PLOS ONE, 13(3), e0194889.
Pinto, J. M., & Castle, J. L. (2022). Machine Learning Dynamic Switching Approach to Forecasting in the Presence of Structural Breaks. Journal of Business Cycle Research, 18(2), 129–157.
Siami‑Namini, S., Tavakoli, N., & Namin, A. S. (2019). The performance of LSTM and BiLSTM in forecasting time series. IEEE Big Data, 3285–3292.
Stempień, D., & Ślepaczuk, R. (2025). Hybrid Models for Financial Forecasting: Combining Econometric, Machine Learning, and Deep Learning Models. arXiv preprint, arXiv:2505.19617.
Sun, X., Yuan, P., Yao, F., Qin, Z., Yang, S., & Wang, X. (2024). Financial Fragility in Emerging Markets: Examining the Innovative Applications of Machine Learning Design Methods. Journal of the Knowledge Economy. DOI: 10.1007/s13132-023-01731-w.
Tokgöz, N., Değirmenci, A., & Karal, Ö. (2024). Machine learning-based classification of turkish music for mood-driven selection. Journal of Advanced Research in Natural and Applied Sciences, 10(2), 312-328.
Wang, J., & Li, X. (2023). Hybrid ARIMA-LSTM models for financial time series: A meta-analysis. Applied Intelligence, 53(6), 7021–7041.
Zhang, Y., & Wang, L. (2023). A comparative study of XGBoost and LSTM for time series forecasting in volatile markets. Expert Systems with Applications, 213, 118876.
