LEVERAGING AI AND MACHINE LEARNING FOR ENERGY-EFFICIENT SMART CITY OPERATIONS

Abstract

The idea behind smart cities is to use advanced technologies for the administration and optimization of the city functionalities, which comprise the energy infrastructure as well. Optimizing energy infrastructure to reduce energy use, costs, and environmental impact is one of the largest challenges smart cities face. The study presents a machine-learning algorithm that accurately forecasts a building's energy efficiency, thereby enhancing the functionality of intelligent cities. The Kaggle Energy Efficiency Dataset is used as the source for the research. Before applying the two most popular regression techniques, Extra Tree and CatBoost, the research performs various data preprocessing steps, such as filling gaps, normalization, and outlier rejection. Evaluation criteria of the model performance include R², RMSE, MSE, MAE, and MAPE. The results show that the models have excellent predictive power, as demonstrated by 99.8% and 99.5% of the R 2 of the ExtraTree and CatBoost model respectively, and the error rate is very low based on all the measures.  The findings demonstrate the superiority of ensemble techniques in handling complex energy consumption patterns; therefore, they are highly applicable to energy efficiency prediction, which would, in turn, streamline smart, data-driven decisions in contemporary cities.