應用或然性分類樹及群集分析於資料庫行銷之研究－以銀行保險代理人公司為例

Probabilistic Classification Tree and Cluster Analysis with Application to Database Marketing －A Case Study of Bancassurance Agency

林聖二(Sheng-Er Lin)

指導教授 ： 羅淑娟

國立臺北科技大學/管理學院/工業工程與管理系所/碩士(2004年)

摘要

本研究之主要目的在使用資料探勘分析工具，對案例銀行保險代理人公司在資料庫行銷上之資料探勘分析，並以案例資料庫來進行實證研究，案例共計23,721筆，在剔除資料不全與流失顧客的資料後，共有15,800筆。
使用資料探勘分析方法包括：以「分類決策樹」分析歸納出尚未購買某一保險商品之潛在客戶群分類法則和目標類別預測結果之機率，並藉此拓展公司和客戶之間的關係；以「分類決策樹」之延伸性，從顧客消費行為中分析出保險商品之間隱藏的依存性關聯規則，找出交叉銷售組合；以「群集分析」協助釐清顧客的相似性，從資料庫中分析出市場區隔。
本案例研究之結果如下：
一、在分析歸納出尚未購買重大疾病健康險之潛在客戶分類法則以及目標類別預測結果之機率中，發現”保險保額”為顧客是否購買該商品最重要的預測子（Predictor），其次為為投保年齡和職業類別。
二、除了分析主附約保險商品之間的依存性關聯規則外，也探索終身壽險、重大疾病健康險以及投資型壽險可以用來選為優先考量的推薦商品，以帶動其它商品的銷售。
三、以群集分析反映顧客之間的相似性中，發現在衡量市場成長策略的情況下，建議以女性之終身壽險和養老壽險與小孩之終身壽險和重大疾病健康險其現有與潛在市場應進行市場滲透與發展。
依據這些資訊在運用資料庫行銷時，就可以將不同保險商品針對不同的顧客，做為潛在市場之滲透與發展。總體而言，本案例研究結果將可提供給經營決策者做為目標潛在客戶開發、交叉銷售組合以及反映市場區隔的決策支援參考方式之一。

關鍵字

資料庫行銷 ； 資料探勘 ； 分類決策樹 ； 關聯規則 ； 群集分析

並列摘要

The main purpose of this study included the understanding of Probabilistic Classification Tree and Cluster Analysis with Application to Database Marketing. The case study of Bancassurance agency established by data mining with probabilistic (Bayesian) classification tree technology and E-Means cluster analysis technology, and that generalizing from the do not yet buy the prospected customer classification rules and the target of predictive results of probability through a classification tree model, and analysing insurance products association rule through extensible classification trees model, and analysing market segmentation through the cluster analysis model.
This study used the data provided by Bancassurance agency to perform the empirical research. The original data was 23,721 records, after deleting the missing data and customer churn data, there were 15,800 records totally. The results were as followed：
1.In addition to doing not yet buy the dread disease insurance target of prospected customer classification rules, it discovered the most predictor that predicted customer to buy was "sum assured", and the insurance age, the insurance occupation.
2.Besides analysis base plan and rider products pattern, it explored whole life insurance, dread disease insurance, and investment insurance to recommend customer with the priority.
3.Clustering customer and analyzing the differences among those clusters, if it will have market growth that suggested the female of whole life insurance, and retirement insurance and the children of whole life insurance, and dread disease insurance marketing are worth penetrating and developing.
The above by analyzing the customer profile in the database system, this case study discovered potential or implied rules and knowledge, and provided the business managements and executives as reference points for Database Marketing which included cross-selling, to develop the target of prospected customer and market segmentation strategies.

並列關鍵字

Database Marketing ； Data Mining ； Classification Tree ； Association Rule

參考文獻

[33] 顏博文，應用資料探勘技術分析學生選課特性與學業表現，中原大學資訊管理學系研究所，碩士論文，中壢，2003。

[39] 周歆凱，利用『資料探勘技術』探討急診高資源耗用者之特性，國立台灣大學醫療機構管理研究所，碩士論文，台北，2004。

Database Marketing,” The Journal of Database Marketing, Vol.6, No.2 1998,

Knowledge Discovery: An Overview”, In advances in knowledge

Discovery and Data Mining, 1996, pp.471-473.