To solve the problem that dispersion of data relations in fishery science data, and the lack of close contact between data and users lead to inaccuracy of the results calculated by traditional recommendation algorithm and make it difficult to recommend suitable fishery science data for users, a recommendation algorithm based on geographic location is proposed. This method introduces geographic location information and forms a cross-domain recommendation algorithm based on the traditional user - article domain. It integrates the geographic location information into recommendation algorithm by calculating the relations between IP and users’ records, and build a complete set of personalized recommendation system for users. The results showed that the recommendation accuracy and average hit rates increased by 9.7% and 3.9% compared to the traditional collaborative recommendation when there were 10 pieces of recommendation data. It makes up the shortcoming that traditional recommendation algorithm does not consider the geographical location information contained in the fishery science data, thus improving the efficiency of recommendation.