Research On Collaborative Filtering Recommendation Based On User Clustering And Latent Factor Model

Due to the progress of science and technology,Internet technology has been rapidly improved.Today,we live in an era of information overload.In order to quickly extract the useful information we need from massive data,we need an intelligent system to make choices for us.It can help us to filter out useful data from huge amounts of data.We call such a system a recommendation system;the core of the recommendation system is a recommendation algorithm,and among the recommendation algorithms,the collaborative filtering recommendation algorithm is well known to us and is often applied to Industrially.However.collaborative filtering has its disadvantages.For example,it has problems such as data sparsity and extensibility.Based on previous research,this paper further studies how to mitigate these problems of collaborative filtering recommendation algorithms.The overall research content of the paper is:First:Improved k-means clustering based on particle swarm algorithm;taking into account the k-means clustering can be automatically clustered by the method of machine learning without specific categories of data,divided into the characteristics of the corresponding k clusters,this article introduces k-means clustering is used to cluster users in the recommendation system,and the users belonging to the same category are divided into the same cluster as much as possible,and the clustering effect of the k-means algorithm is easily considered to be the location of the initial center point.Because of this,the paper considers improving the k-means clustering algorithm by using the PSO algorithm with rapid exploration of global optimal values.Second:Apply the improved k-means clustering algorithm of particle swarm optimization algorithm to the collaborative filtering recommendation algorithm;taking into account that user-based collaborative filtering is to study the similarity among these users within the scope of all users,which makes the user-based The data that the collaborative filtering recommendation algorithm needs to process is very large,which results in the scalability problem of the memory-based collaborative filtering recommendation.In order to alleviate this problem,the improved k-means clustering algorithm of the particle swarm optimization algorithm is added to the collaborative filtering.In the recommendation,the user is clustered,and the calculation range of the similar user is reduced to the user in the user cluster,so the amount of data processed by the algorithm is reduced,the speed of the algorithm recommendation is improved,and it belongs to the same cluster.The users are very similar,so the accuracy of the improved algorithm is also improved.Third:The collaborative filtering recommendation algorithm based on latent factor model and user clustering is proposed.Under the premise of improved collaborative filtering recommendation algorithm based on particle swarm optimization for k-means clustering,an latent factor model is added,considering that it can be used the dimensionality reduction technique is very good at predicting the vacancy value of the user-item rating matrix,thereby more reliably restoring the original sparse user-item rating matrix to a dense user-item rating matrix,so it is introduced and improved.The class algorithm clusters users in a dense user-item rating matrix to improve the accuracy of the clustering,and the recommendation algorithm reduces the calculation range from all users to corresponding clusters when studying the similarity between users.The simulation results of the system verify that the improved recommendation algorithm proposed in this paper is better than the previous recommendation algorithm of particle swarm optimization k-means clustering.