Building a measure to integrate into a hybrid data mining method to analyze the risk of customer

Abstract

Analysis of customer risk is important task in business operation of the enterprise. This paper, first proposes the building of strength measure to rank clusters obtained from clustering algorithm. The next step builds an improved clustering algorithm by integrating of computing this measure into a clustering algorithm. This improved clustering algorithm is named Fuzzy C-Means-Rank (FCM-R). The final step proposes a hybrid data mining method to analyze the risk of the customer. The hybrid method includes two stages. In the first stage, we apply a classifying algorithm to classify customers into two groups: risk and no risk. In the second stage, we apply the improved clustering algorithm FCM-R to cluster risk customers into clusters and to rank clusters by measuring the risk level of clusters. The hybrid method has been applied to a real data set to generate clusters ranked according to the risk level from high to low. With such results, our proposed method will support enterprises to analyze customer risk and to propose appropriate risk management regulation for customers in each cluster.