Personalized recommendation systems have become an important part of e-commerce, social media, and other applications. However, the traditional collaborative filtering algorithm is only based on the user's scoring history of the product, ignoring the attributes and characteristics of the product itself. To solve this problem, this paper proposes a personalized recommendation algorithm based on knowledge graph, which can combine the similarity between goods and user preferences to make recommendations and add the scoring mechanism, thus improving the accuracy and practicability of the recommendation system. Experimental results show that our algorithm outperforms the traditional user-based and item-based co-filtering algorithms in evaluation indexes such as accuracy, recall and F1 value, demonstrating the effectiveness and feasibility of this algorithm in the field of personalized recommendation.