With the explosive growth of Internet information, the recommendation system plays an increasingly important role. In order to solve the problem of sparse information in the traditional recommendation system and to reasonably express the user’s preference, a fuzzy recommendation algorithm (HFR) based on heterogeneous information network is proposed. The HFR algorithm constructs a triangular fuzzy scoring model to fuzzify the user’s discrete scoring information. In addition, it also adds the attribute information of the project and uses the meta-path representation. Based on this, the multi-source information is fully utilized and a new similarity measure is proposed. The score is predicted to get the final recommendation result. The experimental results show that the HFR algorithm effectively solves the problem of sparse information and improves the recommendation quality.