Research On Collaborative Filtering Recommendation Algorithm Based On Tag Clustering And Interest Division

A large amount of information appears in people's vision as the Internet developing rapidly.Information explosion makes it easier for people to receive the messages.But at the same time,the quick acquisition of valuable messages is becoming more discommodious.For avoiding this trouble,people usually retrieve and filter the information before getting it.As a representative of information retrieval technology,search engine can help people to retrieve useful information from the massive information,but when the search keywords cannot reflect the search needs,the results of the search will be disappointing.However,personalized recommendation can just make up for this deficiency,as a typical application of information filtering.At present,the mainstream recommendation methods include content-based recommendation,collaborative filtering recommendation,rule-based recommendation,mixed recommendation and so on.Among them,collaborative filtering recommendation technology is the most widely used in practical application.Collaborative filtering is used to calculate the user's similarity of the target user by the scores of products,then recommending the products to the target user.However,the traditional collaborative filtering algorithm does not consider the impact of the label on the recommendation,mining user interests unilaterally based on the product scores of users,failed to divide the user's interest effectively.At the same time,it ignores the change of user interest over time.In order to solve the above problems,the article has conducted the following study:1.Aiming at the problem that the traditional collaborative filtering algorithm has ignored user interest changes with the time factors,a collaborative filtering recommendation algorithm based on time factor has been proposed in this paper.The algorithm has considered the impact of time weight of the product scores and the degree of interest of the product at different times.It established the time forgetting model and the time window model and merged the two models to generate the time factor.After that,the algorithm filter the product scores using the time factor in the user's similarity calculation to calculate the similar users of target users more accurately and reduce the decline of the quality of the recommended causing by time.Finally,experiments show that this method can adapt to the changes of user interest effectively and improve the accuracy of recommendation in the intelligent web system.2.Considering the relationship between label and user interest,this paper proposes a collaborative filtering recommendation algorithm based on label clustering and interest partitioning.The algorithm considers the influence of labels and product scores of users on the recommended results based on the traditional cooperative filtering algorithm.It classified user interests through label clustering and selected similar users of the target user on the both sides of labels and product scores.And it adds the time factor in the calculation of the weight of the label and product score to adapt to changes in user interest.The experiment shows that the algorithm can divide user interests effectively,reduce the influence of time factor on the quality of recommendation and improve the accuracy of the recommendation with the cross validation and comparison of other recommendation on the Movielens data sets.