ntvНаучно-технический вестник информационных технологий, механики и оптикиScientific and Technical Journal of Information Technologies, Mechanics and Optics2226-14942500-0373Университет ИТМО10.17586/2226-1494-2025-25-6-1150-1159ntv-548Research ArticleКОМПЬЮТЕРНЫЕ СИСТЕМЫ И ИНФОРМАЦИОННЫЕ ТЕХНОЛОГИИCOMPUTER SCIENCEВзвешенная ансамблевая модель, сочетающая ARIMA, LSTM и GBM для надежного прогнозирования временных рядовA weighted ensemble model combining ARIMA, LSTM, and GBM for robust time series predictionhttps://orcid.org/0000-0002-4211-0131ВигнешА.VigneshA.

Вигнеш Арумугам, научный сотрудник

600087; Ченнаи

sc 57219977662

Arumugam Vignesh, Scientific Researcher

600087; Chennai

sc 57219977662

vignesha2@srmist.edu.inhttps://orcid.org/0000-0002-9100-9937ВиджаялакшмиН.VijayalakshmiN.

Виджаялакшми Натараян, кандидат технических наук, доцент

600087; Ченнаи

sc 57212868093

Natarajan Vijayalakshmi, PhD, Associate Professor

600087; Chennai

sc 57212868093

vijayaln@srmist.edu.inИнститут науки и технологий SRM — РамапурамИндияSRM Institute of Science and Technology — RamapuramIndia20252312202525611501159Copyright © Вигнеш А., Виджаялакшми Н., 20252025Вигнеш А., Виджаялакшми Н.Vignesh A., Vijayalakshmi N.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://ntv.elpub.ru/jour/article/view/548

   Прогнозирование временных рядов используется в исследованиях и приложениях в ряде областей, таких как прогнозирование состояния окружающей среды, здравоохранение, финансы, управление цепочками поставок и энергопотребление. Точное прогнозирование будущих значений необходимо для стратегического планирования, эффективности работы и принятия обоснованных решений относительно переменных, зависящих от времени. Предлагается гибридная архитектура прогнозирования временных рядов, сочетающая в себе сильные стороны машинного обучения и статистических моделей, в частности сетей градиентного бустинга (Gradient Boosting Machines, GBM), авторегрессионных интегрированных скользящих средних (Auto-Regressive Integrated Moving Average, ARIMA) и сетей с долговременной краткосрочной памятью (Long Short-Term Memory, LSTM). В то время как сети LSTM и GBM способны улавливать сложные и нелинейные закономерности, модель ARIMA фиксирует линейные компоненты временных рядов. Гибридная модель использует интерпретируемость ARIMA, временную память LSTM и эффективность ансамблевого обучения GBM, интегрируя эти три модели. Комплексные эксперименты, проведенные на эталонных наборах данных, показали, что точность и надежность прогнозов предлагаемой гибридизации значительно превосходят как отдельные модели, так и традиционные базовые. Результаты показывают, что для различных реальных приложений гибридные архитектуры могут обеспечивать надежные и точные прогнозы временных рядов.

   Time series forecasting has been used in research and applications in a number of domains such as environmental forecasting, healthcare, finance, supply chain management, and energy consumption. Accurate prediction of future values is necessary for strategic planning operational efficiency and well-informed decision-making regarding time-dependent variables. A hybrid time series forecasting architecture is proposed that combines the strengths of machine learning and statistical models, in particular Gradient Boosting Machines (GBM), Auto-Regressive Integrated Moving Average (ARIMA), and Long Short-Term Memory (LSTM) networks. While LSTM networks and GBM are able to capture complex dependencies and nonlinear patterns, the ARIMA model captures the linear components within the time series. The hybrid model exploits ARIMA interpretability, LSTM temporal memory ability, and GBM ensemble learning efficiency by integrating these three models. Comprehensive experiments conducted on benchmark data sets have shown that the accuracy and reliability of predictions of the proposed hybridization significantly exceeds both individual models and traditional baseline models. The results show that for a variety of real-world applications, hybrid architectures can deliver reliable and accurate time series predictions.

авторегрессионные интегрированные скользящие сети (ARIMA)долговременная краткосрочная память (LSTM)градиентный бустинг (GBM)прогнозирование временных рядовAuto-Regressive Integrated Moving Average (ARIMA)Long Short-Term Memory (LSTM)Gradient Boosting Machines (GBM)time series forecastingReferencesAghemo C., Blaso L., Pellegrino A. Building automation and control systems: A case study to evaluate the energy and environmental performances of a lighting control system in offices // Automation in Construction. 2014. V. 43. P. 10–22. doi: 10.1016/j.autcon.2014.02.015Aghemo C., Blaso L., Pellegrino A. Building automation and control systems: A case study to evaluate the energy and environmental performances of a lighting control system in offices. Automation in Construction, 2014, vol. 43, pp. 10–22. doi: 10.1016/j.autcon.2014.02.015Arumugam V., Natarajan V. Time series modeling and forecasting using Autoregressive Integrated Moving Average and Seasonal Autoregressive Integrated Moving Average models // Instrumentation Mesure Métrologie. 2023. V. 22. N 4 P. 161–168. doi: 10.18280/i2m.220404Arumugam V., Natarajan V. Time series modeling and forecasting using Autoregressive Integrated Moving Average and Seasonal Autoregressive Integrated Moving Average models. Instrumentation Mesure Métrologie, 2023, vol. 22, no. 4, pp. 161–168. doi: 10.18280/i2m.220404Beard E., Marsden J., Brown J., Tombor I., Stapleton J., Michie S., West R. Understanding and using time series analyses in addiction research // Addiction. 2019. V. 114. N 10. P. 1866–1884. doi: 10.1111/add.14643Beard E., Marsden J., Brown J., Tombor I., Stapleton J., Michie S., West R. Understanding and using time series analyses in addiction research. Addiction, 2019, vol. 114, no. 10, pp. 1866–1884. doi: 10.1111/add.14643Bernal J.L., Soumerai S., Gasparrini A. A methodological framework for model selection in interrupted time series studies // Journal of Clinical Epidemiology. 2018. V. 103. P. 82–91. doi: 10.1016/j.jclinepi.2018.05.026Bernal J.L., Soumerai S., Gasparrini A. A methodological framework for model selection in interrupted time series studies. Journal of Clinical Epidemiology, 2018, vol. 103, pp. 82–91. doi: 10.1016/j.jclinepi.2018.05.026Chou J.-S., Tran D.-S. Forecasting energy consumption time series using machine learning techniques based on usage patterns of residential householders // Energy. 2018. V. 165. Part B. P. 709–726. doi: 10.1016/j.energy.2018.09.144Chou J.-S., Tran D.-S. Forecasting energy consumption time series using machine learning techniques based on usage patterns of residential householders. Energy, 2018, vol.165, part B, pp. 709–726. doi: 10.1016/j.energy.2018.09.144Chujai P., Kerdprasop N., Kerdprasop K. Time series analysis of household electric consumption with ARIMA and ARMA models // Proc. of the International MultiConference of Engineers and Computer Scientists (IMECS). 2013. V. 1. P. 1–6.Chujai P., Kerdprasop N., Kerdprasop K. Time series analysis of household electric consumption with ARIMA and ARMA models. Proc. of the International MultiConference of Engineers and Computer Scientists (IMECS), 2013, vol. 1, pp. 1–6.Corry E., Pauwels P., Hu S., Keane M., O’Donnell J. A performance assessment ontology for the environmental and energy management of buildings // Automation in Construction. 2015. V. 57. P. 249–259. doi: 10.1016/j.autcon.2015.05.002Corry E., Pauwels P., Hu S., Keane M., O’Donnell J. A performance assessment ontology for the environmental and energy management of buildings. Automation in Construction, 2015, vol. 57, pp. 249–259. doi: 10.1016/j.autcon.2015.05.002Dey P., Chaulya S.K., Kumar S. Hybrid CNN-LSTM and IoT-based coal mine hazards monitoring and prediction system // Process Safety and Environmental Protection. 2021. V. 152. P. 249–263. doi: 10.1016/j.psep.2021.06.005Dey P., Chaulya S.K., Kumar S. Hybrid CNN-LSTM and IoT-based coal mine hazards monitoring and prediction system. Process Safety and Environmental Protection, 2021, vol. 152, pp. 249–263. doi: 10.1016/j.psep.2021.06.005Hu Y., Peng L., Li X., Yao X., Lin H., Chi T. A novel evolution tree for analyzing the global energy consumption structure // Energy. 2018. V. 147. P. 1177–1187. doi: 10.1016/j.energy.2018.01.093Hu Y., Peng L., Li X., Yao X., Lin H., Chi T. A novel evolution tree for analyzing the global energy consumption structure. Energy, 2018, vol. 147, pp. 1177–1187. doi: 10.1016/j.energy.2018.01.093Li K., Huang W., Hu G., Li J. Ultra-short term power load forecasting based on CEEMDAN-SE and LSTM neural network // Energy and Buildings. 2023. V. 279. P. 112666. doi: 10.1016/j.enbuild.2022.112666Li K., Huang W., Hu G., Li J. Ultra-short term power load forecasting based on CEEMDAN-SE and LSTM neural network. Energy and Buildings, 2023, vol. 279, pp. 112666. doi: 10.1016/j.enbuild.2022.112666Lee Y.-S., Tong L.-I. Forecasting energy consumption using a grey model improved by incorporating genetic programming // Energy Conversion and Management. 2011. V. 52. N 1. P. 147–152. doi: 10.1016/j.enconman.2010.06.053Lee Y.-S., Tong L.-I. Forecasting energy consumption using a grey model improved by incorporating genetic programming. Energy Conversion and Management, 2011, vol. 52, no. 1, pp.147–152. doi: 10.1016/j.enconman.2010.06.053Liboschik T., Fokianos K., Fried R. tscount: An R package for analysis of count time series following generalized linear models // Journal of Statistical Software. 2017. V. 82. N 5. P. 1–51. doi: 10.18637/jss.v082.i05Liboschik T., Fokianos K., Fried R. tscount: An R package for analysis of count time series following generalized linear models. Journal of Statistical Software, 2017, vol. 82, no. 5, pp. 1–51. doi: 10.18637/jss.v082.i05Zhao N., Zhang H., Yang X., Yan J., You F. Emerging information and communication technologies for smart energy systems and renewable transition // Advances in Applied Energy. 2023. V. 9. N 100125. doi: 10.1016/j.adapen.2023.100125Zhao N., Zhang H., Yang X., Yan J., You F. Emerging information and communication technologies for smart energy systems and renewable transition. Advances in Applied Energy, 2023, vol. 9, pp. 100125. doi: 10.1016/j.adapen.2023.100125Peleg R., R. Weiss, A. Hoogi A. Leveraging the triple exponential moving average for fast-adaptive moment estimation // arXiv. 2023. arXiv:2306.01423v1. doi: 10.48550/arXiv.2306.01423Peleg R., R. Weiss, A. Hoogi A. Leveraging the triple exponential moving average for fast-adaptive moment estimation. arXiv, 2023, arXiv:2306.01423v1. doi: 10.48550/arXiv.2306.01423Pérez-Lombard L., Ortiz J., Pout C. A review on buildings energy consumption information // Energy and Buildings. 2008. V. 40. N 3. P. 394–398. doi: 10.1016/j.enbuild.2007.03.007Pérez-Lombard L., Ortiz J., Pout C. A review on buildings energy consumption information. Energy and Buildings, 2008. vol. 40, no. 3, pp. 394–398. doi: 10.1016/j.enbuild.2007.03.007Pomorski P., Gorse D. Improving on the Markov-switching regression model by the use of an adaptive moving average // Springer Proceedings in Business and Economics. 2023. P. 17–30. doi: 10.1007/978-3-031-23844-4_2Pomorski P., Gorse D. Improving on the Markov-switching regression model by the use of an adaptive moving average. Springer Proceedings in Business and Economics, 2023, pp. 17–30. doi: 10.1007/978-3-031-23844-4_2Mehtab S., Sen J. Stock price prediction using CNN and LSTM-based deep learning models // Proc. of the International Conference on Decision Aid Sciences and Application (DASA). 2020. P. 447–453. doi: 10.1109/DASA51403.2020.9317207Mehtab S., Sen J. Stock price prediction using CNN and LSTM-based deep learning models. Proc. of the International Conference on Decision Aid Sciences and Application (DASA), 2020, pp. 447–453. doi: 10.1109/DASA51403.2020.9317207Saleti S., Panchumarthi L.Y., Kallam Y.R., Parchuri L., Jitte S. Enhancing forecasting accuracy with a moving average-integrated hybrid ARIMA-LSTM model // SN Computer Science. 2024. V. 5. N 6. P. 704. doi: 10.1007/s42979-024-03060-4Saleti S., Panchumarthi L.Y., Kallam Y.R., Parchuri L., Jitte S. Enhancing forecasting accuracy with a moving average-integrated hybrid ARIMA-LSTM model. SN Computer Science, 2024, vol. 5, no. 6, pp. 704. doi: 10.1007/s42979-024-03060-4Xue S., Chen H., Zheng X. Detection and quantification of anomalies in communication networks based on LSTM-ARIMA combined model // International Journal of Machine Learning and Cybernetics. 2022. V. 13. N 10. P. 3159–3172. doi: 10.1007/s13042-022-01586-8Xue S., Chen H., Zheng X. Detection and quantification of anomalies in communication networks based on LSTM-ARIMA combined model. International Journal of Machine Learning and Cybernetics, 2022, vol. 13, no. 10, pp. 3159–3172. doi: 10.1007/s13042-022-01586-8Thangarajan B., Nagaraja M.S., Dhandra B.V. A novel hybrid model for time series forecasting using artificial neural network and autoregressive integrated moving average models // Trends in Mathematics. 2024. Part F2357. P. 747–754. doi: 10.1007/978-3-031-41420-6_62Thangarajan B., Nagaraja M.S., Dhandra B.V. A novel hybrid model for time series forecasting using artificial neural network and autoregressive integrated moving average models. Trends in Mathematics, 2024, part F2357, pp. 747–754. doi: 10.1007/978-3-031-41420-6_62He W., King M., Luo X., Dooner M., Li D., Wang J. Technologies and economics of electric energy storages in power systems : Review and perspective // Advances in Applied Energy, 2021. V. 4. P. 100060. doi: 10.1016/j.adapen.2021.100060He W., King M., Luo X., Dooner M., Li D., Wang J. Technologies and economics of electric energy storages in power systems : Reviewand perspective. Advances in Applied Energy, 2021, vol. 4, pp. 100060. doi: 10.1016/j.adapen.2021.100060Xu W., Peng H., Zeng X., Zhou F., Tian X., Peng X. Deep belief network-based AR model for nonlinear time series forecasting // Applied Soft Computing. 2019. V. 77. P. 605–621. doi: 10.1016/j.asoc.2019.02.006Xu W., Peng H., Zeng X., Zhou F., Tian X., Peng X. Deep belief network-based AR model for nonlinear time series forecasting. Applied Soft Computing, 2019, vol. 77, pp. 605–621. doi: 10.1016/j.asoc.2019.02.006Yip S.-C., Wong K., Hew W.-P., Gan M.-T., Phan R.C.W., Tan S.-W. Detection of energy theft and defective smart meters in smart grids using linear regression // International Journal of Electrical Power & Energy Systems. 2017. V. 91. P. 230–240. doi: 10.1016/j.ijepes.2017.04.005Yip S.-C., Wong K., Hew W.-P., Gan M.-T., Phan R.C.W., Tan S.-W. Detection of energy theft and defective smart meters in smart grids using linear regression. International Journal of Electrical Power & Energy Systems, 2017, vol. 91, pp. 230–240. doi: 10.1016/j.ijepes.2017.04.005Young J.M., Stacey I., Dobbins T.A., Dunlop S., Dessaix A.L., Currow D.C. Association between tobacco plain packaging and Quitline calls: A population-based, interrupted time-series analysis // Medical Journal of Australia. 2014. V. 200. N 1. P. 29–32. doi: 10.5694/mja13.11070Young J.M., Stacey I., Dobbins T.A., Dunlop S., Dessaix A.L., Currow D.C. Association between tobacco plain packaging and Quitline calls: A population-based, interrupted time-series analysis. Medical Journal of Australia, 2014, vol. 200, no. 1, pp. 29–32. doi: 10.5694/mja13.11070Zhou S., Wu Z., Li J., Zhang X.-P. Real-time energy control approach for smart home energy management system // Electric Power Components and Systems. 2014. V. 42. N 3–4. P. 315–326. doi: 10.1080/15325008.2013.862322Zhou S., Wu Z., Li J., Zhang X.-P. Real-time energy control approach for smart home energy management system. Electric Power Components and Systems, 2014. vol. 42, no. 3–4, pp. 315–326. doi: 10.1080/15325008.2013.862322Gidon J.S., Borah J., Sahoo S., Majumdar S. Neural network approaches for enhanced landslide prediction: A comparative study for Mawiongrim, Meghalaya, India // Natural Hazards. 2025. V. 121. N 3. P. 3677–3699. doi: 10.1007/s11069-024-06948-9Gidon J.S., Borah J., Sahoo S., Majumdar S. Neural network approaches for enhanced landslide prediction: A comparative study for Mawiongrim, Meghalaya, India. Natural Hazards, 2025, vol. 121, no. 3, pp. 3677–3699. doi: 10.1007/s11069-024-06948-9Gidon J.S., Borah J., Sahoo S., Majumdar S., Fujita M. Bidirectional LSTM model for accurate and real-time landslide detection: A case study in Mawiongrim, Meghalaya, India // IEEE Internet of Things Journal. 2024. V. 11. N 3. P. 3792–3800. doi: 10.1109/JIOT.2023.3326203Gidon J.S., Borah J., Sahoo S., Majumdar S., Fujita M. Bidirectional LSTM model for accurate and real-time landslide detection: A case study in Mawiongrim, Meghalaya, India. IEEE Internet of Things Journal, 2024, vol. 11, no. 3, pp. 3792–3800. doi: 10.1109/JIOT.2023.3326203Arumugam V., Natarajan V. Enhanced time series forecasting using hybrid ARIMA and machine learning models // Indonesian Journal of Electrical Engineering and Computer Science. 2025. V. 38. N 3 P. 1970–1979. doi: 10.11591/ijeecs.v38.i3.pp1970-1979Arumugam V., Natarajan V. Enhanced time series forecasting using hybrid ARIMA and machine learning models. Indonesian Journal of Electrical Engineering and Computer Science, 2025, vol. 38, no. 3, pp. 1970–1979. doi: 10.11591/ijeecs.v38.i3.pp1970-1979Nadzir M.S.M., Rabuan U., Ali S.H.M., Borah J., Majumdar S., Rohmad M.S. Drone-based air quality monitoring: Development and evaluation of low-cost PM2.5 sensor for remote environmental assessment // Sensors and Materials. 2025. V. 37. N. 6. P. 2153–2171. doi: 10.18494/sam5449Nadzir M.S.M., Rabuan U., Ali S.H.M., Borah J., Majumdar S., Rohmad M.S. Drone-based air quality monitoring: Development and evaluation of low-cost PM2.5 sensor for remote environmental assessment. Sensors and Materials, 2025, vol. 37, no. 6, pp. 2153–2171. doi: 10.18494/sam5449

The authors declare that there are no conflicts of interest present.