Recommendation Algorithm Based On Convolution Neural Network

Internet era has brought explosive growth of data volume,when people shopping online,browsing the Internet,evaluating the goods on the Internet and sharing through social network,generate a lot of data in the server,which contains a lot of important information.In order to make better use of data generated when people are online,spot patterns and recommend suitable products and information to the people more accurately,to solve the problem of people lost in the huge amount of the goods,the accuracy and the performance real real-time of the recommendation system is becoming increasingly important.In the e-commerce website,they can recommend products that users are most interested in by intelligent recommendation,while increasing sales business;in the news media website,they can recommend the most interesting news to users to enhance the news hits;in music as well as books and other sites of interest,you can recommend to the user what they are most interested in to enhance the user experience etc.With the explosive growth of data up recently,how to extract the most accurate information from these data quickly,is related to the major website and their business interests.Research on recommendation system,more importantly,the study of recommendation algorithm aroused widespread interest in the academic and business community.In this paper,we make a deep understanding and research on the concept of recommendation system,the problems and the current popular recommendation algorithm,including some of the common problems in recommended system,such as cold start,comparing the advantages and disadvantages of each recommending algorithm,the problem of data matrix sparsity,common evaluation indicators and so on.After studying and analysing the current status of the recommendation system,with the combination of my prior knowledge of graphics,I apply the convolution neural network to the recommended system and achieve the recommendation algorithm based on convolution neural networks,in the case of to maintaining the recommended accuracy,we greatly reduce the time of the traditionalrecommendation algorithm based on depth learning.In the course of the study,combining with the feature of convolution neural network and data set characteristics,thinking about how to represent and quantitative data set and construct the network model,selecting the dimensions of the the feature,we realize the recommendation algorithm based on convolution neural network in Matlab.We use a publicly available data set that is MovieLens-1M,to do experiment,comparing with the traditional recommendation algorithm and achieving good results in the experiment.