Research On Collaborative Filtering Algorithm Based On Tag And Trust Relationship

The personalized recommendation technology actively recommends the items that may be of interest to users,which is an effective means to alleviate the problem of information overload,and is of great significance for the development of the Internet in the era of big data.Collaborative filtering is the most widely used and most successful recommendation algorithm.It has received extensive attention in academia and industry,and obtained some achievements.However,there is still exists some problems so that the recommendation accuracy is not high due to data sparsity.Aiming at the sparseness of data of collaborative filtering algorithm,this thesis will study the related optimized algorithms in order to obtain better recommendation quality.The main research contents are as follows: 1.For the most existing collaborative filtering methods use ratings when calculating the item similarity,but ignore the item tag information,we proposed a new collaborative filtering recommendation algorithm combining ratings and tag.Using item tag model and item popularity to calculate the correlation between items,and linearly weights with the similarity of ratings.Experimental results show that the improved algorithm is better than some existing improved algorithms.2.In the view of the fact that the collaborative filtering algorithms generate comprehensive interest based on the ratings,and neglect the emphasis of user preferences,an optimization collaborative recommendation algorithm for tag interest model based on ratings correction is proposed.The algorithm analyzes the user's interest by user tag frequency and user tag ratings,and uses the user tag interest model to improve the similarity calculation method.The experimental results show that the proposed algorithm not only improves the accuracy of ratings prediction,but also increases the accuracy of recommendation to some extent.3.Aiming at the problem that the existing collaborative filtering algorithm only considered the user similarity or trust relationship when selecting the user neighborhood,but ignored the similarity of emphasis of user preference and asymmetry of the trust relationship,an improved collaborative filtering algorithm combining tag and trust value is proposed.The algorithm calculates the user similarity by using the user tag interest model,and constructs the asymmetry of the trust relationship using the number of user ratings,and selects the neighbor users according to the similarity between users and the user trust value.The experimental results show that the optimized algorithm improves the accuracy of the recommendation and relieves the cold start problem to some extent.