The present invention discloses a cooperative caching method based on feature fusion. The method employs base station data sets of a mobile Internet to obtain a content similarity matrix and a location similarity matrix of users. The user records include base stations and Internet access contents accessed respectively by different users at different times; the content similarity matrix and the location similarity matrix are fused to obtain a similar matrix after the fusion; the similarity matrix after the fusion is used to cluster the users to obtain the users' category relations; the users' category relations are mapped to the base stations to obtain the cooperation relations of the base stations; caching strategies of the base stations are calculated according to the obtained cooperation relations of base stations. The method makes use of the base station data sets of the mobile Internet to propose a new method to determine the caching content of the base station from the perspective of cooperation of base stations.