Research On Dynamic Recommendation Of User Behavior Based On Data Fusion

Under the background of big data information age,with the rapid development of the Internet,the amount of information generated by the major e-commerce platforms every day shows an exponential "explosive" growth trend.Behind the information,there is a huge immeasurable value,which can only be used by businesses and users after a series of transformations,so as to realize its true value.In the face of huge and complex information,how to help users obtain the information they need conveniently and efficiently and how to accurately judge whether the screened information is really needed by users has become a challenging work.As one of the application fields of artificial intelligence,to a large extent,the emergence of recommendation system solves the above problems.Personalized recommendation system can carry out personalized recommendation according to different needs of different users.Among them,collaborative filtering recommendation algorithm is the most widely used recommendation algorithm,which mainly uses the user's rating matrix to predict the target user's rating of other unknown projects.However,in the era of big data with explosive growth of data,the number of users and project resources of each major e-commerce platform is only increasing.In the face of such a huge project resource,each user only has rating information for a small part of them,and different users have evaluated different projects,which makes the problem of data sparsity increasingly significant.At the same time,a single user rating information can not fully describe the user's real interest preferences,and the problem of user data information fragmentation is significant.In the face of the above situation,it is more and more difficult for users to select projects of real interest.The recommendation quality of the recommendation system based on traditional collaborative filtering is significantly reduced,so new improved methods are needed to solve the recommendation problem.K-means clustering algorithm is a common clustering algorithm.It can process the data in the early or middle stage,divide all the data into different classes according to certain rules,so that all kinds of divided data can be "similar within the class,different between the classes".In this paper,K-means clustering algorithm and collaborative filtering are integrated,and an improved collaborative filtering recommendation model based on K-means clustering is proposed.The integration of the two reduces the time of calculating similarity and improves the efficiency of recommendation by reducing the search scope of the target user's nearest neighbors.At the same time,considering that the independent and single user rating information is not enough to accurately describe the user's interest preference,this paper combines the attribute feature information of the project,constructs the user project feature relative rating matrix,studies the further construction of user interest preference model,and proposes a dynamic user behavior recommendation model based on data fusion.After the model is established,experiments are designed to observe the influence of different clustering number,different similarity measurement methods and different number of neighbors on the model.Through the comparative analysis of the experimental results,the optimal parameter value of the algorithm is obtained,and then the final algorithm New-HBCF is obtained.Finally,four evaluation indexes of MAE,Precision,Recall and F1 are selected to evaluate.The experimental results show that under the verification of these four evaluation indexes,the proposed dynamic recommendation method based on data fusion is indeed superior to the traditional UBCF and the existing two improved algorithms,which proves that validity of the New-HBCF algorithm proposed in this paper.In the last chapter,this paper make the conclusion of the research,and puts forward some supposing for the future research,and expects that the existing research in this paper can provide some reference and help for the future research in the field of recommendation algorithm.