On identifying the irregularities of electricity customer behaviors using soft computing approach

Abstract

This study aims to implement a soft computing approach in identifying the irregularities of customer behaviour on the use of prepaid electricity pulses. The used methods are Support Vector Machine (SVM), Naïve Bayes (NB), Classification and Regression Tree (CART), and k-Nearest Neighbours (KNN). To evaluate the performance of the classification system, a 10-fold Cross Validation technique is used for the historical data of pre-paid pulse purchase transactions. Validation results are measured using accuracy, precision and recall values. This research shows that all soft computing methods gave good performances in classifying electrical consumption behaviours. CART method has the highest accuracy value of 100% compared with others algorithm. At precision values, KNN and CART methods have the highest precision value among other algorithms that are 99% to 100%. Whereas, the recall values of each method has a high recall value of 100%. Moreover, each method can predict morbidity accurately because the addition of the amount of data testing does not affect the performance of each method.