OPTIMIZATION OF FORECASTING THROUGH INCORPORATION OF CAUSAL INFLUENCE

DOI: 10.31673/2786-8362.2025.011761

Abstract

The article explores approaches to improving forecasting accuracy by incorporating causal
influence between variables with time delays. The study uses Granger causality to detect temporal lags and
adjust input variables accordingly. A method is proposed to enhance LSTM-model performance by
integrating lag analysis into the training pipeline. Experimental evaluation shows that accounting for
different lag times across variables significantly reduces forecasting error. The results confirm that causallag modeling improves robustness and reflects real-world dynamics in time series forecasting. The authors
emphasize the importance of considering lagged interrelationships to form a more informative input space
for the neural network. Experimental results indicate an improvement in forecasting accuracy when causally
significant relationships are identified in advance.
Keywords: forecasting, time series, Granger causality, lag effect, prediction model, neural network

References
1. Habibnia A., Etesami J., Kiyavash N. Modeling Systemic Risk: A Time-Varying
Nonparametric Causal Inference Framework. SSRN. 2024. URL:
https://doi.org/10.2139/ssrn.4684230.
2. Research on condition operation monitoring of power system based on supervisory control
and data acquisition model / B. Li et al. Alexandria Engineering Journal. 2024. Vol. 99. P. 326–334.
URL: https://doi.org/10.1016/j.aej.2024.05.027.
3. Open-Source Internet of Things-Based Supervisory Control and Data Acquisition System for
Photovoltaic Monitoring and Control Using HTTP and TCP/IP Protocols / W. Khalid et al. Energies.
2024. Vol. 17, no. 16. P. 4083. URL: https://doi.org/10.3390/en17164083.
4. Lopes F. M., Dutra E., Boussetta S. Evaluation of Daily Temperature Extremes in the
ECMWF Operational Weather Forecasts and ERA5 Reanalysis. Atmosphere. 2024. Vol. 15, no. 1.
P. 93. URL: https://doi.org/10.3390/atmos15010093.
5. Siddiqi A. Value Analytics for Earth Observing Systems. IGARSS 2024 - 2024 IEEE
International Geoscience and Remote Sensing Symposium, Athens, Greece, 7–12 July 2024. 2024.
P. 3864–3867. URL: https://doi.org/10.1109/igarss53475.2024.10642411.
6. Liao S.-h., Widowati R., Lee C.-Y. Data mining analytics investigation on TikTok users'
behaviors: social media app development. Library Hi Tech. 2022. URL: https://doi.org/10.1108/lht08-2022-0368.
7. Chiang L. H., Braatz R. D. Process monitoring using causal map and multivariate statistics:
fault detection and identification. Chemometrics and Intelligent Laboratory Systems. 2003. Vol. 65,
no. 2. P. 159–178. URL: https://doi.org/10.1016/s0169-7439(02)00140-5.
8. Nadim K., Ragab A., Ouali M.-S. Data-driven dynamic causality analysis of industrial
systems using interpretable machine learning and process mining. Journal of Intelligent
Manufacturing. 2022. URL: https://doi.org/10.1007/s10845-021-01903-y.
9. Беспала О.М. Інструментарій причинно-наслідкового висновку: огляд та
перспективи. Control Systems and Computers. 2020. № 5 (289). С. 52–63.
URL: https://doi.org/10.15407/csc.2020.05.052.
10. Deep understanding in industrial processes by complementing human expertise with
interpretable patterns of machine learning / A. Ragab et al. Expert Systems with Applications. 2019.
Vol. 122. P. 388–405. URL: https://doi.org/10.1016/j.eswa.2019.01.011.
11. Беспала О. М., Отрох С. І.. Ружинський В. Г.. Моделювання спрямованого ациклічного
графа для причинного висновку. Наукові записки Державного університету телекомунікацій.
2023. № 2 (1) С. 87-95 DOI: 10.31673/2518-7678.2023.020202.
12. Беспала О. М. Метод пошуку та оцінки впливу причинно-наслідкових зв’язків в
системах прийняття рішень. Вісник НТУ "ХПІ". Серія: Інформатика та моделювання.
2020. № 2 (4). С. 59 – 72.
13. Granger C. W. J. Investigating Causal Relations by Econometric Models and Cross-spectral
Methods. Econometrica. 1969. Vol. 37, no. 3. P. 424. URL: https://doi.org/10.2307/1912791.