Collaborative Filtering Recommendation Algorithm Based On Improved K-means Clustering Algorithm And RBM

With the development of the Internet and the popularization of the Internet,massive information came into being and the world entered an era of big data.In the face of dizzying and overloaded information,how to obtain useful information more effectively and find the information that interests us has become a new and urgent problem.When people take the initiative to find information,the search engine can solve the user's needs to a certain extent,but it is not the best solution.On the one hand,when the keywords can not be accurately described,the search results will be greatly reduced;on the other hand,Can not satisfy people's desire for passive access and push information purposes.The proposed system provides a revolutionary solution to this problem.It can be recommended algorithm based on the user's historical behavior analysis,the establishment of the model,which proactively pushed to the user interested in the information.Recommended system has been widely used in various fields,such as e-commerce site's product recommendation,music and movie recommendations,news and so on.On the one hand,the application of this technology can improve the user experience of the system and increase the user's stickiness to the system;on the other hand,it can also attract more users.The core of a recommendation system is based on what kind of recommendation algorithm is used.The research in this field mainly focuses on the improvement and optimization of the recommendation algorithm.At present,the mainstream recommendation algorithms include content-based recommendation algorithms,collaborative filtering-based recommendation algorithms,association rule-based recommendation algorithms and hybrid recommendation algorithms.Among them,the recommendation algorithm based on collaborative filtering is one of the most widely used and most popular algorithms,Which uses the user's history to score the project to generate the project scoring matrix and calculates the similarity of users or projects for recommendation,but there is data sparseness,which can not meet the requirements of cold start and scalability of new users and new projects.Needs and development of the Internet.With the development and improvement of machine learning technology,the combination of traditional recommendation system and machine learning also greatly enhances the recommendation effect.This article in order to solve the above problems,made some improvements,including two aspects:First of all,the traditional K-means algorithm uses the Euclidean distance in calculating the similarity,which has drawbacks,resulting in the inaccurate clustering result.This paper introduces trust relationship and time attenuation in the process of calculating the similarity between users,which not only considers the influence of scoring,but also considers the change of trust relationship and interest,which can get better and more realistic clustering results.Find the nearest neighbor with the highest similarity according to the clustering result,and carry out the weighted calculation of predictive score according to the similarity.Secondly,the RBM model of the constrained Boltzmann machine has achieved good results on the collaborative filtering problem.In this paper,the RBM model and the improved K-means algorithm are used to predict the results.On the one hand,the improved K-means algorithm is used to cluster Users are divided into different clusters.When predicting the scores,the cluster is judged first,which can reduce the computational load to some extent and improve the efficiency of the algorithm.On the other hand,the RBM algorithm considers only the user's rating of the project and does not consider the difference between users The similarity between the two algorithms can improve the accuracy of the algorithm.