Based on the samples of 605 patent infringement cases judged during the year 2002 to 2012, the empirical study is conducted to explore the impact factors of compensation about the patent infringement damage in china by using the multiple regression analysis model. The results show that 12 variables, such as ‘the time of infringement’, ‘the number of infringers’ variables have a significant effect on the amount of damage judged. Secondly, by selecting the above 12 variables and establishing a BP neural network model, the study forecasts the amount of damage. The results show that by applying this model, the accuracy of forecasting for compensation on IP infringement ranging from RMB 0-50,000 can reach 94%; the accuracy of forecasting for compensation on IP infringement ranging from RMB 50,000-100,000 can reach 68%; the accuracy of forecasting for compensation on IP infringement ranging from RMB100,000-500,000 can reach 85%. Finally, some recommendations are given to improve patent compensation system in china.