To the kinds of automobile insurances, the revenue of physical damage insurance always keep a high ratio. According to the research, he premium of automobile physical damage insurance is ten times of automobile third party liability insurance. On the other side, the change of automobile physical damage insurance's loss ratio can affect the other automobile insurance. That is to say, whether the automobile physical damage insurance can earn profit or not will affect the business performance of property insurance company. The study apply the automobile physical damage insurance's underwriting data which comes from a domestic representative property insurance company. Furthermore, we use logit model, ANN, GANN model and neuro-fuzzy model in the underwriting management. Empirical results show that the neuro-fuzzy can outperform all models. The results can be provided as practical underwriting application.