A Tag Recommendation Algorithm For Mixed Memory Curve And Association Rules

With the development of information era and the massive growth of data,users and resources data in application system become very huge.It is becoming difficult for users to find the information they need.The tag has the function of expressing the users' tastes and marking the features of the resources.Therefor users can use labels to search,filter,classify and annotate resources.It can increase the user's experience feeling and retention of the system by recommending appropriate tags to the users.The factors that affect the accuracy of the label recommendation include the following two points?The first one is the problem of drifting user interest.Users' preferences will drift as time goes on and the environment changes.This phenomenon will make negative effect that the quality of the label recommendation is declining.The second is the problem of data sparsity.In large scale system,the amount of users and resources' data are very huge that will cause the probability of intersection between different users decreases,and then lead to the data sparsity phenomenona.It will create a bad influence on the system diging users' preferences.For the sake of overcoming the two problems that talking above,the paper proposes the following corresponding solutions.Firstly,aiming at the problem of drifting user interest,a Tag Recommendation algorithm based on Memory Curve is proposed.In order to portray user preferences more accurately and detect user interest changes in time,the algorithm improves the Collaborative Filtering based on the principle of memory curve.The second,aiming at the phenomenon of data sparsity,a Tag Recommendation algorithm based on Association Rules is designed in this paper.This method defines an overlapping sliding time window used to collect users' label data.And mining association rules between labels is based on that data,then recommend the label according to the association rules.Finally,to improve the quality of tag recommendation,a Tag Recommendation algorithm for Mixed Memory Curve and Association Rule is proposed.The algorithm combines the above two label recommendation algorithms in a weighted average way,and uses Particle Swarm Optimization algorithm to find the best weighting parameters.The experimental results show that the Tag Recommendation algorithm based on Memory Curve can accurately describe the user preferences,improve the recommendation accuracy.The Tag Recommendation algorithm based on Association Rules can alleviate the data sparseness problem in a certain extent,improve the quality of recommendation.And the Tag Recommendation algorithm for Mixed Memory Curve and Association Rule can effectively integrate the above two methods,and improve the accuracy of recommendation.