先進分類模式於保費數據分析之應用

Application of Advanced Classification Model for Insurance Fee Data Analysis

成熟自由的金融保險市場，琳琅滿目的儲蓄理財工具，「保險」逐漸成為國人解決風險帶走擔憂的重要規劃，始的近年投保率攀升。目前家庭收入爬升緩慢，國人生計憂心，出現了續期保費繳不出的窘境。保險公司重要營運來自於保費收入，客戶保費續繳問題為保險公司與業務員關注議題。本研究經發達的3C系統與電腦科技結合，採用保險業界客戶保險資料，透過資料探勘選出19個條件屬性1個決策屬性。屬性選取技術，運用K鄰近法、規則、貝氏分類與決策樹四種演算法，執行保費續繳問題之預測，並找出:繳費人、薪資結構、購買保單數與有效保單數，為影響續期保費繳納之重要因子。實驗證明屬性選取後準確率高，決策樹J48為較優的演算法，研究結果提供給業界參考，達到公司、客戶與業務員三贏最優局面，為本研究之貢獻。

The mature and free financial and insurance market, as well as a wealth of savings and wealth management tools, "insurance''has gradually become an important plan for Chinese people to solve risks and take away worries. The insurance rate has risen in recent years. At present, the family income has been climbing slowly, and the country is worried about its livelihood. The important operation of insurance companies comes from premium income, and the issue of customer premium renewal is a topic of concern for insurance companies and salesmen. In this study, the developed 3C system is combined with computer technology, using customer insurance data from the insurance industry, and 19 condition attributes and 1 decision attribute are selected through data exploration. Attribute selection technology, using K proximity method, rules, Bayesian classification and decision tree four algorithms to perform the prediction of premium renewal problem, and find out: payer, salary structure, number of purchased policies and effective number of policies, as the impact Important factor for renewal premium payment. Experiments prove that the accuracy of attribute selection is high, and the decision tree J48 is a better algorithm. The research results are provided to the industry for reference, to achieve the best situation of win-win for the company, customers and sales staff, and contribute to this research.