Implementation of SVM Kernels for Identifying Irregularities Usage of Smart Electric Voucher

Abstract

Statistical methods and machine learning have been widely used to identify deviations in the use of electrical energy for prepaid services. The paper applies the Support Vector Machine method to identify prepaid electricity usage irregularities that can overcome classification and regression problems with linear or nonlinear kernels with high accuracy and relatively small error rates. The results showed that the predictions of morbidity of electricity voucher purchase transactions, the amount of test data used did not affect the accuracy, precision, and memory values of the Linear and Polynomial kernels, the values obtained were all 100%. This shows that the addition of test data, the value of False Positive and False Negative remains 0. Thus, in each additional test data value of precision, accuracy and memory do not change. However, in the RBF kernel, the value of accuracy and precision decreases as the amount of test data increases.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint