La détection des anomalies dans la consommation d’électricité en utilisant des méthodes de détection des outliers

Abstract

The goal of this project is to use time series anomaly detection methods to identify electrical failures and wasted energy in residential buildings. In this context, we used three methods from unsupervised learning to detect anomalies in electricity consumption in buildings. These methods are: Forest Isolation, One-Class SVM, and K-Means. The results obtained showed that each method estimates such a data as an anomaly or not and that the method based on One-Class SVM showed its performance compared to the others.