Personalized Clothing Recommendation Based on Knowledge Graph

摘要

Most clothing recommender systems employed the approach of collaborative filtering and content-based which pay more attention to predicting the user's item preferences based on massive historical data. These systems can easily lead to the dilemma of cold start and data sparseness, furthermore, they overlooked user's context, such as weather, occasion, requirements, emotions and also correlations between clothing items. In this paper, we construct knowledge graph of user, knowledge graph of clothing and knowledge graph of context, utilize Apriori algorithm to capture the intrinsic correlations between clothing attributes and context attributes. Thereafter, we search and match the established knowledge graph according to the user's requirements and combine the Top-N algorithm to generate the recommendation results directly. In the meantime, we recommend which has the most similar to user's collected clothes through the similarity of clothing ontology to improve the efficiency and accuracy of the recommendation. The experimental results show that our method can effectively solve the cold start problem and has higher recommendation quality compared with traditional collaborative filtering.