With the rapid development of internet and global positioning technology, location-based social network (LBSN) is emerging in large numbers which encourages users to share their personal feelings and locations in real time by check-ins. Volumes of check-in data afford an opportunity for mining user preference, which promotes location-based services such as point of interest (POI) recommendation. POI recommendation can not only help users identify favorite locations, but also help POI owner acquire more target customers. A location's category is the accurate abstraction of the context semantics of location. Most of present research only directly considers user preference on a specific location and ignore consideration of location's category. In Yelp, we find the ratio of common visited location is lower than that of common visited location category, which means that considering user preference on location category is more reasonable than that on specified locations. In light of the above, we present a novel POI recommendation method based on location category and social network named CSRS which infers users' preference on category from their check-ins history, and at the same time take the differences of category preferences among friends into consideration. The experimental results on Yelp demonstrate CSRS achieves superior precision and recall compared to other recommendation techniques.