Application of Collaborative Filtering Optimization Algorithm Based on Semantic Relationships in Interior Design

Due to the diversity of interior design, it is difficult for users to mine target data, so personalized recommendation systems for users are particularly important. Therefore, an optimized collaborative filtering recommendation system is proposed. Firstly, a random walk recommendation model based on category combination space is constructed, abandoning the traditional flat relationship connection and using Hasse diagram to achieve one-to-one mapping between items and types. The semantic relationship and distance are defined. Finally, a basic recommendation framework for random walks is established based on data such as jump behavior. Next, the potential semantic relationships between entities are explored, and a lightweight knowledge graph is proposed to define the social and explicit relationships between entities. Finally, the short-term features of the project are obtained using deep collaborative filtering technology, and a deep collaborative filtering temporal model based on semantic relationships is constructed. In subsequent validation, these experiments confirmed that under the vector dimension of 10, the average HR@K and NDCG@K were 6.9% and 12.9% higher than the other models. Therefore, the collaborative filtering recommendation model based on semantic relationships proposed in the study is reliable.