Due to the large amount of training data and the high complexity of its recommend algorithm, the updating cycle of recommendation system tend to be long. However, the data on the system is growing all the time, and a lot of data is produced during the cycle, which is useful for the recommendation of next moment, and recommendation system can't use these data in time. In order to use these data in time to improve the quality of recommendation system, a new approach to hybrid recommendation based on incremental data was proposed. The approach mainly divided recommendation into offline and online module, the offline module is used to produce the personalized recommendation list, while the online recommendation module maintains a list of popular trend momentum based on real-time and incremental data. Then, combining with the results of the two modules, based on which give users anonymous or personalized recommendation. Experiments show that the approach is simple, effective, feasible, and can improve the performance of recommendation system better.