Analysis of Household Electricity Consumption Segmentation using K-means Clustering

Abstract

This research focuses on predicting household electricity usage by applying K-Means Clustering segmentation in support of energy-saving strategies. In this study, we gather historical monthly electricity consumption data over three years for analysis, considering attributes such as the number of occupants and the number of electrical appliances. The segmentation process resulted in three main clusters: low, medium, and high consumption. This segmentation enables easier identification of consumption patterns and serves as a foundation for constructing more accurate and targeted prediction models. The prediction model was developed using both linear and non-linear (exponential) regression methods. Evaluation results show that the non-linear model delivers the best performance, with a correlation of up to 99.84% and lower error values compared to the linear model. The integrative approach combining clustering and prediction proves effective in identifying consumption characteristics and supporting adaptive and sustainable decision-making in household energy efficiency management.