The paper introduces a novel approach called HERec for heterogeneous information network (HIN) based recommendation. HINs are used to model complex and heterogeneous auxiliary data in recommender systems, but existing methods often rely on path-based similarities, which may not fully capture the latent structure features of users and items. To address this, HERec proposes a meta-path based random walk strategy to generate meaningful node sequences for network embedding. The learned node embeddings are then transformed by a set of fusion functions and integrated into an extended matrix factorization (MF) model. Extensive experiments on three real-world datasets demonstrate the effectiveness of HERec, showing its capability in handling cold-start problems and improving recommendation performance. The key contributions of the paper include a novel HIN embedding method guided by meta-paths and a flexible fusion approach to integrate different embeddings.The paper introduces a novel approach called HERec for heterogeneous information network (HIN) based recommendation. HINs are used to model complex and heterogeneous auxiliary data in recommender systems, but existing methods often rely on path-based similarities, which may not fully capture the latent structure features of users and items. To address this, HERec proposes a meta-path based random walk strategy to generate meaningful node sequences for network embedding. The learned node embeddings are then transformed by a set of fusion functions and integrated into an extended matrix factorization (MF) model. Extensive experiments on three real-world datasets demonstrate the effectiveness of HERec, showing its capability in handling cold-start problems and improving recommendation performance. The key contributions of the paper include a novel HIN embedding method guided by meta-paths and a flexible fusion approach to integrate different embeddings.