Personalized Social Image Tag Recommendation Algorithm Based on Tensor Decomposition

This paper proposes a three-dimensional tensor model, which uses the three dimensions of the three-dimensional tensor to describe three types of entities in the social tag recommendation system, and builds a recommendation model for three different data sets. First introduced the traditional label recommendation algorithm and its shortcomings, and then introduced the personalized social speech image recommendation algorithm based on tensor decomposition. The user performs similarity calculation and classifies users with similar characteristics to ensure that the data has initial aggregation and optimized the weight. Experiments show that this algorithm reduces the data sparsity problem in the tag recommendation system and improves the accuracy of the recommendation results.