科技的進步促使產業往前發展，而金融（Finance）與科技（Technology）的碰撞，讓金融創新展開了ㄧ個新的競爭起點與激盪出新的思維。保險公司若可藉此改善客戶之服務、著重風險之管理、並提升經營之效率，將可能藉此邁入頂尖保險公司之列。在資訊的普及及國人保險意識的抬頭下，保險商品的資訊取得已較以往迅速及便利，客戶在購買保險商品上也更精打細算，不易受到保險業務員的話術影響購買保險商品時的意願。另一方面，在個人化的行動裝置普及下，在客戶進行購買行為時，保險公司也更容易收集到大量個人化差異的交易資訊。因此傳統的保險行銷策略及技術均受到衝擊，現代化的保險公司皆已捨去任由業務員大量行銷及產品差異化行銷，輔助業務員尋找最可能銷售成功之保險商品，轉向為鎖定客戶市場區分進行目標行銷。藉由目標行銷保險公司也可以根據客戶需求，將商品區分不同的通路區隔，在不同的銷售通路提供不同的組合商品或是專案價格，以求更接近目標市場。而藉由收集交易資料集進行資料採擷便是達成此目的的重要環節之一。
本研究以某保險公司承保資料為研究對象，利用資料探勘之技術找出成功銷售保險商品時客戶特徵之關聯性。資料探勘中的群集分析演算法常用於分析市場區隔，將有類似特徵的消費者區分為若干群組。本研究探討Cascade K-means演算法及商品特性區分客戶族群之差異，亦即進行顧客特徵分類，再利用HotSpot關聯演算法擷取每一群集之特徵關聯性，期能得知資料潛藏之意義。目標為找到交易資料集中客戶購買保險商品的特徵關聯性，並且發掘保險商品的市場區隔。研究的最終目標為讓保險公司可對每一客戶族群之特性加以分析與了解，依其屬性訂定專屬之行銷策略、保險商品、或培訓相關策略之專業人才。

Technological evolution drives the development of industry. When technology and the financial industry are brought together, it inspires creative ideas and new competition in financial innovation. For insurance companies, the introduction of FinTech may help improve customer service, strengthen risk management, elevate management efficiency, or even push the company into the list of top insurers. With the convenient availability of information and increased public awareness of insurance concepts in Taiwan, knowledge about insurance products can be easily and quickly acquired. Customers are becoming more price-sensitive when purchasing related products, and less likely to be convinced by insurance agents. On the other hand, the prevalence of personal mobile devices makes it easier for insurance companies to acquire a large amount of differentiated transaction information from individual customers when they purchase insurance policies. Such changes have impacted the traditional marketing strategies and techniques of the industry. Insurance companies nowadays have to give up the conventional methods of soliciting business, where mass marketing and product differentiation are encouraged to help insurance agents identify potential “best sellers.” Instead, insurers tend to concentrate their efforts on target marketing, trying to focus on selected customer segments. Through target marketing, insurance companies can also arrange different distribution channels for different products based on customer demand. Variant product combinations or price packages are offered to different sales channels so as to better accommodate the needs of the target market. Data collection and subsequent data mining, no doubt, are important steps to achieve these purposes.
This study is targets the underwriting information of an insurance company. Using data mining techniques, the research endeavors to explore the relationship between customer characteristics and the insurance products which have been sold successfully. The cluster algorithms adopted herein are technique frequently used to analyze market segmentation, where consumers with similar characteristics are grouped together. For this research, Cascade K-means algorithm is used to classify the customers, taking into consideration product attributes. In other words, customers are first grouped based on their characteristics. The relationship among the clusters are then analyzed using the HotSpot Algorithm, so as to probe the underlying significance of this data. These approaches are taken to explore the relation of customer characteristics for those insurance products with transaction data, and in the meantime identifying market segmentation for different products. The ultimate objective of this research is to help the insurance company analyze and understand the unique features of each customer group, provide tailor-made marketing strategies and insurance products, or develop professionals that are able to fulfill related strategies based on these attributes.

摘要 I
目次 IV
表目錄 VI
圖目錄 VII
第一章 緒論 1
1.1 研究背景 1
1.2 研究動機 2
1.3 研究目的 3
1.4 研究流程 4
1.5 論文架構 5
第二章 文獻探討 6
2.1 臺灣地區人身保險介紹 6
2.2 資料探勘探討 17
2.3 相關研究探討 27
第三章 研究方法 28
3.1 研究設計與架構 28
3.2 資料來源介紹 30
3.3 研究方法選擇 31
3.4 研究工具介紹 33
3.5 資料前處理 36
第四章 資料分析與解釋 43
4.1 料探勘分析 43
4.2 研究結果解釋 56
第五章 結論 57
5.1 研究結論 57
5.2 研究限制 58
5.3 未來研究方向與建議 59
參考文獻 60

一、英文文獻：
Berry, M. J., & Linoff, G. (1997). Data mining techniques: for marketing, sales, and customer support. John Wiley & Sons, Inc..
Caliński, T., & Harabasz, J. (1974). A dendrite method for cluster analysis. Communications in Statistics-theory and Methods, 3(1), 1-27.
DAVIES D L, BOULDIN D W. (1979). A cluster separation measure[J]. IEEE transactions on pattern analysis and machine intelligence, 1979, PAMI-1(2): 224-227. DOI:10.1109/TPAMI.1979.4766909.
DIMITRIADOU E, DOLNICˇAR S, WEINGESSEL A. (2002). An examination of indexes for determining the number of clusters in binary data sets[J]. Psychometrika, 2002, 67(1): 137-159. DOI:10.1007/BF02294713.
Fayyad, U.; Piatetsky-Shapiro, G..; Smyth, P. (1996). “From Data Mining to Knowledge Discovery: An overview”, In advances in Knowledge Discovery and Data Mining, Fayyad et al, pp.471-493, 1996.
Friedman, J.H. and Fisher, N.I. (1999). Bump Hunting in High-dimensional Data, Statistics and Computing, Vol. 9, Issue 2, 123-143.
Gasparino, C. (2000). “Goldman stakes claim to future of big board in spear deal,” Wall Street Journal, New York, N.Y., Eastern Edition, 2000.
Geoffrey Hinton, Terrence J. Sejnowski（editors，1999） Unsupervised Learning and Map Formation: Foundations of Neural Computation, MIT Press, ISBN 0-262-58168-X.
Hall, M., Eibi, F., Holmes, G., Pfahringer, B., Reutemann, P. (2009). Witten, I.H.: The WEKA Data mining Software: An Update. SIGKDD Explor. 11(1), 1018 (2009).

J. B. MacQueen (1967): "Some Methods for classification and Analysis of Multivariate Observations", Proceedings of 5-th Berkeley Symposium on Mathematical Statistics and Probability, Berkeley, University of California Press, 1:281-297.
KRZANOWSKI W J, LAI Y T. (1988). A criterion for determining the number of groups in a data set using sum-of-squares clustering[J]. Biometrics, 1988, 44(1): 23-34. DOI:10.2307/2531893.
Langhnoja, S. G., Barot, M. P., & Mehta, D. B. (2013). Web usage mining using association rule mining on clustered data for pattern discovery. International Journal of Data Mining Techniques and Applications, 2(01).
Lent, B., Agrawal, R., & Srikant, R. (1997). Discovering trends in text database. Proceedings of the 3rd International Conference on Knowledge Discovery and Data Mining.
Bach, M. P., & Varga, M. (2005, January). Creating profile of data mining specialist. In 27th International Conference on Information Technology Interfaces.
Mitchell, T. M. (1997). Machine learning. WCB.
Poitras, E. G., Lajoie, S. P., Doleck, T., & Jarrell, A. (2016). Subgroup Discovery with User Interaction Data: An Empirically Guided Approach to Improving Intelligent Tutoring Systems. Journal of Educational Technology & Society, 19(2), 204-214.
Poitras, E. G., Lajoie, S. P., Doleck, T., & Jarrell, A. (2016). Subgroup Discovery with User Interaction Data: An Empirically Guided Approach to Improving Intelligent Tutoring Systems. Journal of Educational Technology & Society, 19(2).
Sung, S. F., Hsieh, C. Y., Yang, Y. H. K., Lin, H. J., Chen, C. H., Chen, Y. W., & Hu, Y. H. (2015). Developing a stroke severity index based on administrative data was feasible using data mining techniques. Journal of clinical epidemiology, 68(11), 1292-1300.
XIE X L, BENI G. (1991). A validity measure for Fuzzy clustering[J]. IEEE transactions on pattern analysis and machine intelligence, 1991, 13(8): 841-847. DOI:10.1109/34.85677.

二、 中文文獻：
Han-Chul Park（2003），「新行銷通路帶來的挑戰」，壽險季刋，第 128 期。
王秀茹（2009），應用資料探勘於客戶線上購買保險行為之分析，中國文化大學，碩士論文。
朱明媛（2018），高級職業學校圖書館於校園位置與周圍設施對圖書館使用之關聯研究。
吳瑞川（2004），資料探勘技術應用於保險業-以壽險保障為例，東吳大學，碩士論文。
林群弼（2002），保險法論，第 548-551 頁，台北，三民書局。
許舒博（2012），「中華民國一○一年度人壽保險業概況」，中華民國人壽保險商業同業公會年報，台北。
游杰（2001），「保險電子商務網住保險版圖」，現代保險雜誌，第 139 期，34-36 頁，。
劉介傳（2013），利用資料探勘技術探討台灣壽險市場發展之研究，嶺東科技大學，碩士論文。
潘維大、范建得、 羅美隆著（2005），商事法，第 393-394 頁，台北，三民書局，2005 年修訂 7 版。
蔡博清（2006），「我國壽險業通路發展策略之研究--以業務員及銀行保險通路為例」，中正大學，碩士論文。
謝耀龍（2004），壽險行銷，三版，華泰文化。
簡禎富、許嘉裕 (2014) ，資料挖礦與大數據分析。新北市：前程文化。
三、網路資源：
財團法人保險事業發展中心(2012)，「中華民國台灣地區人壽保險重要統計資料」，台北。https://www.tii.org.tw/opencms/information/information1/000001.html
陳勇汀（2017），Association Rule Mining with Specific Right-Hand-Side: HotSpot Algorithm in Weka
http://blog.pulipuli.info/2017/08/wekahotspot-association-rule-mining.html
陳勇汀（2017）， Determin the Optimal Number of Clusters: Cascade K-means
http://blog.pulipuli.info/2017/10/k-determin-optimal-number-of-clusters.html
陳鍾誠 (2010年04月27日)，(網頁標題) 陳鍾誠的網站首頁，(網站標題) 陳鍾誠的網站，取自 http://ccckmit.wikidot.com/main ，網頁修改第 4 版。
Hitachi Vantara (2018) , Mark Hall (2010) , HotSpot Segmentation-Profiling
http://wiki.pentaho.com/display/DATAMINING/HotSpot+Segmentation-Profiling
HotSpot - Weka
http://weka.sourceforge.net/doc.packages/hotSpot/weka/associations/HotSpot.html
Weka 3: Data Mining Software in Java
https://www.cs.waikato.ac.nz/ml/weka/
Machine Learning Group at the University of Waikato. (2016). Weka 3 - Data Mining with Open Source Machine Learning Software in Java. Retrieved July 20, 2017, from http://www.cs.waikato.ac.nz/ml/weka/