The Study of Decision Tree and Artificial Neural Network in Personal Accident Claims

• Main Authors: Chun-Hao Chen, 陳俊豪
• Other Authors: Chin-Peng Chu
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/50137662499282635537

碩士 === 國立東華大學 === 企業管理學系 === 97 === German scholar Farny ever said “It is always that someone will misuse systems created by human being to do evils no matter in which era. So is the insurance system. Insurance fraud is one of more dangerous criminals people could think of.” Because of its low risk, high profit and very little chance to be prosecuted for its criminal liability, it happens in an endless stream. It is difficult to estimate the actual amount of criminals, so it will not be a fair claim when insurance fraud occurs. Insurance fraud increases the loss ratio of insurance company. All policy holders need to pay higher. Governor is very concerned of the getting serious problem of insurance fraud, so establishes “The Insurance Anti-Fraud Institute of R.O.C.” in 2004 to prevent insurance fraud criminals. However, due to limited budget, there is no formal data of insurance fraud. Every Non-Life & Life Insurance Association has to self-help itself developing an index of fraud detection. Personal Accident Insurance has characters of low insurance premium, high protection, no need of physical examination and being effective immediately, so it is easy to fall as an object of insurance fraud. Non-Life Insurance business volume has been increased continuously ever since accident insurance starts. But some non-life insurance companies are using automobile insurance claim staff in their organization to handle accident claims. This research wishes to find the major variant affecting moral hazard, to provide non-life insurance company automobile insurance claim agent a clear understanding of each case and to shorten the educational training time. This research finds the major variant is on underwriting document. Insurant’s sex, age & profession are variants of claiming documents. Across segmentation consult for doctor, insured company numbers, days in hospital, get file day, claim reasons, hospital information are visible variants on claiming documents. But this research is decision reference for underwriting and claiming of insurance business only.