類神經網路應用於擬定汽車保險費率

汽車保險是與消費者關係最為密切的財產保險，但或許因為國人對汽車保險的認知不足，至今仍存在不合理現象。例如：近年汽車車體損失險的投保率下降且損失率逐年上升，其原因或可歸咎於現行的保費不見得反映實際的風險，但此有違精算費率精神的現象若持續下去，勢必對汽車保險的財務健全有不良影響。本文採用國內某產險公司1999年至2002年汽車車體損失保險資料，探討保費收入與理賠支出的關係，希冀在滿足保費均衡的原則下，尋求較小變異數的預測方法，以降低風險。本文考量過去用於產險的最小誤差估計法，以及根據經驗建構模型的類神經網路法，比較這兩種方法何者較能降低分類的誤差與縮小個體的誤差，以期保費收入與理賠支出兩者間有較小的差異。實證結果顯示，現行國內車體損失險不完全符合保費均衡原則，其間仍存在保險補貼。而在模型配適上，最小誤差估計法計較能改善收支不平衡的現象；而類神經網路法的加減費系統具有較大加減幅度，更能有效區分高低風險群組，降低不同危險群組間的補貼現象，並在跨年度的資料中具有較小的誤差變異。

關鍵字

汽車車體損失保險；最小誤差估計法；類神經網路

並列摘要

Among all casualty insurances, auto insurance is the most common in our daily lives. Despite its popularity, there exists awkward phenomenon in auto insurance business. For example, the insured rate of Automobile Material Damage Insurance is going downward but the loss ratio is climbing upward. By charging corresponding premium based on individual risks, we can attract low risk entrants and collect reasonable premiums from the highly risk groups. To further illustrate the concept, we aim to take Automobile Material Damage Insurance for example, to study the most efficient estimator of the future claim. In this study, we compare Minimum Biased Estimate, a previously used method, and Artificial Neural Network (ANN). The reason for including the ANN is due to the fact that the relationship of loss experience (input) and future claim estimation (output) is similar to how the human brain performs. The comparison is based on achieving the minimum error of classes or individuals, using the data between 1999 and 2002 from an insurance company. We found that cross subsidization exists in Automobile Material Damage Insurance. In addition, the new rate produced by minimum bias estimate can alleviate the unbalance between the premium and loss. However the ANN classification rating can allocate those premiums more fairly, where 'fairly' means that higher premiums are paid by those insured with greater risk of loss and vice-versa. Also, the ANN is more efficient than the minimum bias estimator in the panel data.