The Clustering for Clients in a Bank Based on Big Data

Many technologies about data mining, such as clustering, have been widely applied in the context of bank to understand the behaviors of the clients and investors. Unlike some classic clustering validity index using compactness and separation, we employ Pairing Frequency Clustering Validity Index (PFCVI), which uses pairwise pattern information and focuses more on logical reasoning than geometrical features. We use PFCVI to evaluate the clustering quality under different c and find the optimal value of c is 11 based on the bank’s data, and clients in the 11 classes have different savings value potential levels and different fluctuation patterns. Then, we sum up the above 11 classes into 5 categories with different fluctuation patterns – stabilized savings value category, fluctuating savings value category, rising savings value category, falling savings value category and abnormal savings value category. Finally, we analyze each category with techniques like user profile and give some targeted advice for each category aimed at optimal market segment.