Segmentasi Dan Klasifikasi Perilaku Pembayaran Pelanggan Pada Perusahaan Multimedia Dengan Algoritma K-Means Dan C4.5

Abstract

A service provider company often have problem with insolvent customer. As a company that can not verify customer financial, a service provider company is having difficulty on identifying solvent customer. As one of the service provider company in Jakarta which offer internet connectivity and cable TV service, identifying solvent customer is important because, every customer inside coverage area is a potential customer to subscribe their product. In a related research, algorithm  C4.5  become  one  of  the  favorite  algorithm  to  use  in  classifying customer insolvency, and was able to produce  high accuracy model, and easy to understand. The major problem of algorithm C4.5 is attribute selection. As noted that a service provider company cannot verify customer financial, they have no data about customer financial which is one important attribute in credit scoring. Therefore, customer payment behavior will be extracted as customer potential as one of classification attribute. Feature extraction with k-means was able to improve the accuracy of the algorithm C4.5. As noted by improve on accuracy from 59.02% to 77.31% and AUC from 0.537 to 0.836. The customer potential attribute can also be used as reference in the promotion, retention, and managing insolvent customer.