應用資料探勘於銀行顧客行銷分析

Application Data Mining in bank customer marketing analysis.

高文信(Wen-Hsin kao)

指導教授 ： 羅惠瓊

淡江大學/商管學院/企業管理學系碩士在職專班/碩士(2019年)

https://doi.org/10.6846/TKU.2019.00645

摘要

在現今競爭的金融環境下，各銀行為了增進收益，除了重視新客源開發外，也開始重視既有資源的運用；但是如何從既有客戶資源中，挖掘出適當的客戶來銷售合適的產品，用來縮短尋找客戶的時間，並提高成交機率，值得探討。簡單的說，即是從既有交易資料中，找出業務人員、產品及客戶的相關性，藉由將客戶分類，再找出每類型客戶高成交機率的相關產品。
本研究以X銀行2017年及2018年個金業務銷售理財產品資料為研究範圍，銷售筆數共9,448筆，資料包含產品資料、業務資料及客戶資料等變數18項，經資料前置處理，整理出符合本研究範圍的資料 2,033筆，並將資料導入資料探勘中決策樹CHAID演算法進行驗證及分析。
本研究結果顯示，依客戶風險屬性值分類後，各對應的產品正確百分比，訓練樣本為82.8%，檢定樣本為83.3%，足證本研究有相當高的準確度。故透過資料探勘決策樹CHAID演算法對客戶進行分類，再依分類後所對應可銷售的商品資料，可做有效的配置及利用既有客戶資源，大幅降低業務人員找尋客戶的時間成本及提高銷售商品的成交機率，進而增進公司收益。

關鍵字

資料探勘 ； 決策樹 ； 交叉驗證

並列摘要

In today's competitive financial environment, in order to increase revenue, banks have begun to pay attention to the use of existing resources in addition to paying attention to new customer development.
However, how to dig out appropriate customer to sell suitable products from existing customer resources. It is worthwhile to explore the time to find customer and increase the chance of trading. To put it simply, it is to find out the relevance of business personnel, products and customers from the existing transaction data, and to classify the customers, and then find out the relevant products with high transaction probability for each type of customers.

This study takes X Band's 2017-2018 personal financial business sales wealth management product data as the research scope, with a total of 9,448 sales. The data includes 18 items of productdata, business data and customer data. 2,033 pieces of data were obtained in accordance with the scope of this study, and the data was imported into the Decision tree CHAID algorithm in data exploration for verification and analysis.

The results of this study show that, according to the classification of customer risk attribute values, the correct percentage of each corresponding product, the training sample is 82.2%, and the verification sample is 83.3%. The evidence of this study has quite high accuracy. Therefore, though the data exploration Decision tree CHAID algorithm to classify customers, and then according to the saleable commodity data after classification, can effectively configure and utilize existing customer resources, greatly reducing the time cost and sales of business personnel looking for customers. The probability of the transaction of the goods, thereby increasing the bank's income.

並列關鍵字

Data Mining ； Decision Tree ； Cross Verification

參考文獻

參考文獻

中文部分

1. 宋明哲(1996)，「保險學：純風險與保險」，五南書局。

2. 沈清正、陳彥良、陳仕昇、高鴻斌、張元哲、陳家仁、黃琮盛(2002)，「資料間隱含關係的挖掘與展望」，資訊管理學報。

3. 林傑斌(2002)，「資料採掘與OLAP理論與實務」，文魁書局。