Neighborhood based models are widely used in recommender systems nowadays; they are very important and practical but suffering from the time-consuming calculation of similarities based on user feedback.Recent studies have found the combination of the feedback and external data will largely improve the performance and relieve the sufferings.To further leverage more types of data,researchers suggest to make recommendation in heterogeneous information networks context.However,most of these studies have not noticed the differences between user feedback and other types of relations,which could cause severe problems in accuracy as well as running time when importing new data.In this paper,we propose a novel model called Heter-Neighbor,where feedback and external data are treated separately.We firstly provide a new schema of the heterogeneous information network in recommender systems (HIN-RECSYS) to identify new neighborhoods without time-consuming calculations and then integrate the new neighborhoods with models factorizing the similarity matrix into low rank vectors,which are used to make prediction after global optimization.According to empirical studies,our approach outperforms several stateof- the-art models in handling with heterogeneous data.