Research And Implementation Of Video Content Hybrid Recommendation Algorithm Based On Big Data

Since China put forward the "Internet plus" national strategy,the Internet technology has witnessed a rapid development,whether it is our way of shopping and entertainment have become more convenient and intelligent,data and information content showed a rapid increase,how to dig out valuable information is also a heat from these large data The scientific research project and recommendation system can not only provide users with excellent product experience,but also bring great economic benefits to enterprises.Personalized recommendation algorithm is more and more applied to all walks of life,and with the complexity of the specific business scene,there are many excellent solutions in the field of recommendation algorithm.This paper deeply studies the popular recommendation algorithm,and expounds the improved mixed recommendation algorithm,and has made a better conclusion.Fruit.The main research work of this article is as follows:1.There are still some problems in the current recommendation system.For example,the accuracy of the recommendation is not high,the data sparsity is serious,and the single machine model and the existing processing technology are also difficult to meet the demand of mass data processing.In this paper,the popular recommendation algorithm is studied in depth,and an improved hybrid recommendation algorithm is formed by optimizing the combination,and good results are achieved.2.In depth research and Study on collaborative filtering algorithm,the emphasis is on the research and analysis based on singular value decomposition(SVD)and spark MLlib ALS(Alternating Least Squares)recommendation algorithm.The sparse and high-dimensional user project matrix is reduced,and the RMSE(Root Mean Squared Error,RMSE)index is introduced in the matrix reconstruction and decomposition,in order to improve the accuracy of the prediction score,and finally two A linear combination of secondary prediction results is used to get the final recommendation results.3.Introducing Hadoop and spark principles and technology,design and build a multi node distributed architecture cluster.With one of the computer as the main Master node and the other four computers as Slave nodes,the hybrid film recommendation algorithm system is designed and implemented on the spark cluster,and the powerful algorithm library of spark is used to provide better scalability andease of use for the whole system.It improves the data storage performance and processing speed of the system.4.Through the proposed algorithm designed in this paper,the MovieLens data set is successfully used to experiment and verify,the contrast mechanism of different dimensions is adopted,and after many strict experimental results are compared and analyzed,the operation time of the five cluster mode is 60% higher than that of the single machine environment.Compared with the single machine mode,the precision of the recommendation is improved in the distributed cluster nearly 4%.In this paper,based on the hybrid recommendation algorithm on spark,this paper improves and optimizes the precision,data sparsity,scalability and other indicators.