Abstract
Research on energy consumption using machine learning techniques has become increasingly prevalent in recent years due to the growing interest in improving energy efficiency, optimizing energy usage, and promoting sustainability. The purpose of this study is to explore energy consumption trends in Kazakhstan using machine learning algorithms to obtain effective forecasts. Algorithms such as Decision Tree, Random Forest, and KNN were evaluated using cross-validation. Random Forest emerged as the most effective model, achieving the lowest Mean Squared Error (MSE) of 122,077.62 and an R² score of 0.8985±0.0199. KNN displayed accuracy between 0.77 and 0.82, with MSE ranging from 109,470.71 to 257,093.29, while Decision Tree's performance varied significantly with train-test splits, yielding MSE values of 162,418-313,635 and R² scores of 0.724-0.873. Models performed best with a 90%- 10% train-test split, highlighting the importance of training data size for optimal predictive accuracy.