EXPLORING MACHINE LEARNING TECHNIQUES FOR SEASONALITY DETECTION IN RESOURCE BALANCING SYSTEMS

The article investigates machine learning methods for detecting seasonality in resource balancing systems. Modern algorithms (XGBoost, LSTM, TFT, Prophet) are analyzed. A comparative evaluation of methods is conducted based on accuracy (MAE, RMSE), noise resistance, interpretability, and computational complexity. Results show that the optimal method depends on data characteristics, accuracy requirements, and available resources. Recommendations for hybrid approaches to improve forecasting efficiency are proposed.