Research On Personalized News Recommendation System Based On Spark Platform

In the field of news,there are a large number of reading information of the news that users have viewed,so it is especially important to tailor the user's personalized news recommendation.In addition,as the amount of data continues to grow,solving scalability becomes the most important issue,and combining the Spark distributed big data computing platform with the recommendation system can effectively solve this problem.Firstly,the paper summarizes a wide range of recommendation algorithms,particle swarm optimization algorithms,DBSCAN,etc.At the same time,the paper describes in detail the three main components of the distributed computing platform Spark: Spark RDD,Spark MLlib and Spark running framework.The collaborative filtering recommendation algorithm based on LFM implicit semantic model and its parallel implementation are studied.Secondly,a DBSCAN clustering algorithm based on PSO is proposed,which is referred to as PSO-DBSCAN algorithm.The LFM algorithm is deeply analyzed and improved by time function and user similarity calculation function.A fusion recommendation algorithm combining PSO-DBSCAN and improved LFM algorithm is proposed: using time function to user-item interest degree matrix is weighted,and then the LFM model is used to reduce the dimension and fill the missing value processing.Then,the user in the matrix is subjected to PSO-DBSCAN density clustering,and then the user similarity is passed in the category cluster to which the target user belongs.The calculation function finds the k nearest neighbors of the target user,and finally weights the score value of the target user according to the nearest neighbor's score data,and adopts the top-N news recommendation method for recommendation.Parallelization of the fusion recommendation algorithm is implemented on the Spark distributed platform.After that,the data set is used to test and analyze the performance of the proposed hybrid recommendation algorithm on the Spark cluster.It can be seen that the accuracy of the fusion recommendation algorithm is significantly improved compared with the algorithm,and the distributed environment is found.The recommended algorithm execution rate is higher.Finally,a personalized news recommendation system based on the fusion recommendation algorithm is implemented,and the results of news recommendation are displayed.Figure 32;Table 8;Reference 51...