利用类神经模糊理论测寿险保单早期失效之研究.doc

利用類神經模糊理論Predicting Early Lapse of Life Insurance Policies by Using Neuro-Fuzzy Theory 陳建勝(Jian-Shen Chen) 朝陽科技大學保險金融管理系教授 Professor, Department of Insurance, Chaoyang University of Technology 陳美菁*(Mei-Ching Chen) 朝陽科技大學企業管理系副教授 Associate Professor, Department of Business Administration, Chaoyang University of Technology 顏鳳妮Feng-Ni Yan) 朝陽科技大學保險金融管理研究所碩士 Master, Department of Insurance, Chaoyang University of Technology *通訊作者，E-mail：mcchen@cyut.edu.tw Tel: 04轉4202。 利用類神經模糊理論Predicting Early Lapse of Life Insurance Policies by Using Neuro-Fuzzy Theory 摘要 壽險保單早期失效率是業務品質的重要評估指標之一，但隨著壽險市場的開放及我國加入世界貿易組織的影響，壽險公司如何有效地降低保單早期失效率，以提升公司的經營績效，便成為壽險公司的重要課題之一。本研究用類神經模糊理論預測壽險保單效之，其研究變數的選取方式採因素分析及文獻回顧法兩種，並採用整體正確率、型I誤差、型II誤差及誤判成本來衡量預測模式的優劣性。本研究實證結果發現利用因素分析所篩選的變數會比根據文獻回顧所篩選出的變數對保單是否失效的影響大，且類神經模糊在整體上有較穩定之預測能力而利用類神經模糊規則庫，可更具體顯示變之間的相互關連性，被保人年齡及婚姻、要保人婚姻、要保人和被保人關係、是否為員工自保件、年繳化附約保費、附約佔總保費比例對保單是否失效的影響最為顯著。類神經模糊、型I誤差、型II誤差、誤判成本、規則庫The early lapse ratio of the life insurance policies is one of the important estimative indicators. With the opening of life insurance market, it is one of the major lessons to decreasing the lapse ratio to improve the merit of the insurance company. This paper predicts if the life insurance policies is lapse by using Neuro-Fuzzy. There are two ways to select variables: factor analysis and literature review. We use the total actual rate, type I and type II error, and misclassification cost to evaluate the effectiveness of the models. The empirical results show that variables selected by using factor analysis have significant effect on the early lapse of the life insurance policies than by using literature review. In additions, Neuro-Fuzzy model has stably predictive ability, and it describes the relation among variables by using rule blocks of Neuro-Fuzzy. It shows that those variables effect the lapse ratio significantly, including the age, marriage of the insured, the marriage of the applicant, the relation of the insured and applicant, loading pr