Detecting Anomalous Energy Consumption from Profiles

Abstract

Abstract Controlling and reducing electric energy consumption is a critical issue across all over the countries for human wellbeing. However, approaches to achieve energy consumption reduction of individual occupants have not been established yet as both problems of collecting a large amount data of energy consumption and constructing prediction models are challenging. Here we show a case-study of energy consumption analysis, in which households with anomalous energy consumption can be completely detected using a questionnaire about their profiles without seeing actual energy consumption. Our approach is based on simple data mining techniques of outlier detection and decision trees, hence it can be easily implemented in the condominium housing market.