Study on Life Insurance Products Repurchasing Predictive Model with Comparison Among Different Proportions of Oversampling Method

• Main Authors: Chun-Ting Liu, 劉峻廷
• Other Authors: Te-Hsin Liang
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/31285601125988683066

碩士 === 輔仁大學 === 統計資訊學系應用統計碩士班 === 101 === With the market of insurance opened and internationalized. For the entire Taiwan insurance industry, it gets stuck in national and international environment and affects its performance of interests. As these reasons, insurance industry begins to face the problems of foreign capital moving out and company merging. According to some information, it indicates that the insurance industry for every financial institution is needed, and it also shows that 60 percentage insurers will purchase insurances again in the future which means there are 745 thousand people will insure again. Therefore, repurchase insurance market is critical. Thus, this study used data-mining approach to build a predictable model of repurchase. Although there are much studies focusing on obtaining the model, this study used model building and “Oversampling” approach which is used for rare event. This study was proceeded by three oversampling proportion which were 1:1, 1:2 and 1:3. Moreover, some evidence-based researches showed that using 1:1 proportion can get the appropriate repurchased prediction model. In addition, the important factors of affecting consumers repurchase include “age”, “occupation”, “the proportion of insurance to income “, ” have accident insurance or not”.