Collaborative Filtering Recommender Algorithm Based on Ontology and Singular Value Decomposition
Weicheng Liu, Quan Li · 2019
Aiming at the recommendation accuracy of user-item rating matrix sparse generation in e-commerce, we propose a Collaborative Filtering (CF) recommender algorithm based on ontology and Singular Value Decomposition (SVD). Firstly, we build a hierarchy tree based on the ontology concept, and calculate the semantic similarity between items. Then, fill the partial missing values in the user-item sparse matrix according to the semantic similarity, and adjust the threshold so that the intrinsic attribute features of the original matrix are not destroyed. Finally, based on the padding score matrix, the SVD is used to predict and fill the missing values in the sparse matrix again. We test the method taking MovieLens as dataset and compare the experimental results with traditional methods. The result showed that the proposed method effectively mitigates the sparseness of CF recommender and improving its accuracy.