Energy Consumption Prediction Using Random Kitchen Sink (RKS) and Ridge Regression: A Data-Driven Approach

This research introduces a hybrid methodology integrating Random Kitchen Sink (RKS) with Ridge Regression for automated energy consumption prediction. The approach leverages multidimensional data, including temporal features and fuel type information, to forecast energy usage patterns. Our methodology begins with comprehensive feature extraction through data preprocessing and one-hot encoding of categorical variables. The RKS-Ridge model transforms the feature space using random Fourier features to approximate kernel functions followed by Ridge Regression with L2 regularization. Model performance is systematically evaluated against Support Vector Regression (SVR) using metrics including R² score and Mean Squared Error (MSE). Results demonstrate that the hybrid RKS-Ridge approach outperforms traditional SVR, achieving an R² score of 0.036953 compared to SVR's 0.011330, and lower MSE values of 5.886531 versus 6.043150. These findings establish this methodology as a promising tool for energy consumption forecasting, with potential applications in sustainable energy management and demand response strategies.