Aplication of Data Mining for Predictive Analysis of Energy Consumption In Urban Areas for Smart City Development

Abstract

The increasing energy consumption issues in urban areas demand innovative solutions for more efficient and sustainable energy management. This study, conducted in Sukoharjo, Central Java, Indonesia, aims to develop a predictive analysis model for urban energy consumption using data mining techniques within the context of Smart City development. The research method involves collecting and cleaning data from IoT sensors, smart meters, and historical data, followed by the application of clustering techniques, regression, and Random Forest prediction algorithms to build the prediction model. The results indicate that factors such as energy rates, location, time, type of energy users, population density, historical energy consumption, and environmental temperature play significant roles in influencing energy consumption. The predictive model developed using Random Forest performs well, with a Mean Absolute Error (MAE) of 579.10 and a Root Mean Squared Error (RMSE) of 659.71, indicating the model's accuracy in predicting energy consumption. Feature analysis shows that energy rates, district location, and time have the highest importance levels in the prediction model. This research provides valuable insights for energy policy planning in major cities and contributes to the development of more efficient and environmentally friendly Smart Cities.