This paper proposes HERec, a novel heterogeneous information network (HIN) embedding-based approach for recommendation. HINs are used to model complex and heterogeneous auxiliary data in recommender systems, and HERec addresses the challenges of effectively extracting and utilizing information from HINs. The key contributions of HERec include: (1) a meta-path guided heterogeneous network embedding method to capture semantic and structural information of HINs; (2) a general embedding fusion approach to integrate different embeddings based on different meta-paths into a single representation; (3) a novel HIN embedding-based recommendation model, HERec, which effectively integrates various embedding information in HINs to enhance recommendation performance. The model incorporates three flexible fusion functions to transform HIN embeddings into useful information for recommendation. Extensive experiments on three real-world datasets demonstrate the effectiveness of HERec. The results show that HERec outperforms existing methods in terms of rating prediction accuracy, and it is capable of alleviating the cold-start problem. The proposed approach is effective in capturing the latent structure features of users and items, and it provides a more flexible and robust framework for HIN-based recommendation.This paper proposes HERec, a novel heterogeneous information network (HIN) embedding-based approach for recommendation. HINs are used to model complex and heterogeneous auxiliary data in recommender systems, and HERec addresses the challenges of effectively extracting and utilizing information from HINs. The key contributions of HERec include: (1) a meta-path guided heterogeneous network embedding method to capture semantic and structural information of HINs; (2) a general embedding fusion approach to integrate different embeddings based on different meta-paths into a single representation; (3) a novel HIN embedding-based recommendation model, HERec, which effectively integrates various embedding information in HINs to enhance recommendation performance. The model incorporates three flexible fusion functions to transform HIN embeddings into useful information for recommendation. Extensive experiments on three real-world datasets demonstrate the effectiveness of HERec. The results show that HERec outperforms existing methods in terms of rating prediction accuracy, and it is capable of alleviating the cold-start problem. The proposed approach is effective in capturing the latent structure features of users and items, and it provides a more flexible and robust framework for HIN-based recommendation.