Personalized Recommendation (PR) aims to generate user-specific items in response to users' preferences. While existing works have developed various personalized recommendation methods which are effective to some extent, they require setting the explicit and implicit feedback between users and items. In contrast to traditional Collaborative Filtering (CF)-based approaches that suffer from sparse data and cold-start issues, our proposal of mining the higher-order auxiliary information between users and items by combining the knowledge graph (KG) construction with hyper-graph, which is more expressive and semantically plausible. To address these issues, we propose employing hyper-graph to empower knowledge graph construction to express and integrate much richer semantic information for improving the performance of recommendation, namely Hyper-graph based Personalized Recommendation (HPR), for the personalized recommendation task. Our model consists of two components. First, we employ the hyper-graph to model the higher-order relations between users, through the calculation and analysis of the user-item interaction matrix, the users with the highest similarity are seemed as hyper-edges to construct the user hyper-graph. Second, we propose to use knowledge graph to capture and express auxiliary information for recommendation. The two components are integrated in a principled way for returning a more accurate result. Our experimental results on a MovieLens-lM and another Book-Crossing datasets suggest that our proposed HPR framework provides a promising improvement for the personalized recommendation task, and significantly outperforms the state-of-the-art methods. Moreover, the component for mining higher-order relations by hyper-graph is so flexible that can help to improve the performance of KG-based personalized recommendation models significantly