Subspace Frequency Analysis--Based Field Indices Extraction for Electricity Customer Classification

Abstract

In electricity customer classification, the most important task is to avoid the curse of dimensionality problem, as the consumption diagrams have a large number of dimensions. To avoid the curse of dimensionality problem, field indices (load shape factor) are often used instead of consumption diagrams. Field indices are directly extracted from consumption diagrams according to a predefined formula. Previous studies show that the most important thing for defining such a formula is to find meaningful time intervals from consumption diagrams. However, the inconvenient thing is that there are still a lack of details to explain how to define such time intervals. In our study, we propose a data mining--based method named SFATIE to support the extraction of field indices. The performance of the proposed method is evaluated by comparing it with other dimensionality reduction methods during the classification. For the classification, most often we have used classification methods like C5.0, SVM, Neural Net, Bayes Net, and Logistic. The experimental results show that our method is better or close to other dimensionality reduction methods. In addition, the experimental results show that our proposed method can produce the good quality of field indices and that these indices can improve the performance of electricity customer classification.