Abstract
This study focuses on the application of machine learning methods and statistical time series modeling to historical data on natural gas production in Kazakhstan (2000-2024) in order to build a reliable predictive model. The study considers and compares ARIMA, Holt-Winters, linear regression with lag variables, Random Forest, and gradient boosting models. The accuracy of these models was evaluated using standard metrics: MAE, RMSE, and the coefficient of determination R². The comparison results showed that the Holt–Winters exponential smoothing method provides the highest forecast accuracy among all the approaches tested. This model was chosen to generate the forecast of natural gas production volumes for 2025-2027. According to the forecast, a further gradual increase in natural gas production is expected in 2025-2027 while maintaining the identified trend and seasonal patterns. The results obtained demonstrate the effectiveness of integrating modern machine learning algorithms with classical time series analysis methods when working with historical statistical data. The practical significance of this work lies in the fact that the developed forecasting model can contribute to more substantiated strategic planning in the gas industry and improved efficiency of resource management.