An Analysis of Customer Behaviors on Wealth Management of Bank-An Application of Multivariate Statistical Analysis

• Main Authors: Tang, Sheng-Lung, 湯勝隆
• Other Authors: Hsu, Esher
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/8ufnh9

碩士 === 國立臺北大學 === 統計學系 === 103 === Nowadays, the development of the internet and the improvement of the transportation have subverted the business model of the financial industry and have changed our habits to make choices from various financial service providers. Not only from the local banks, people can also get information much easily through various newly development platforms. Decreasing loyalty and dependence also take shape because customers have less connection with the specific financial service provider. These revolutions also make the financial industry more competitive than ever. Banks try to create new marketing strategy to catch more new customers; however, it costs much more to get a new customer than to maintain the old one. As a result, they use information technology - RFM analysis system (Recently, Frequency, Monetary Amount) combine with SOM (Self-Organizing Map) to record and analyze the transaction data from customer and try to find out the most valuable group of customers. Targeting the key group of customer makes the marketing strategy accurate and efficient, thereby increasing customer loyalty and business profit. In this study, 8,636 customers’ transaction data in 2013 have been used to analyze. Multivariate statistical analysis with RFM analysis was used to understand the data. The Two-stage Cluster Analysis and SOM were used to find out the characteristic of the data and further divide customers into four groups with 3 variables in RFM as the benchmark of grouping. By comparing with Two-stage Cluster Analysis, the result of grouping by SOM is much reasonable. Accordingly, a suggestion based upon the study results was provided for bank marketing strategy planning. Keywords: Wealth Management, RFM, Two-stage cluster method, SOM