Research On Collaborative Filtering Algorithm Based On Deep Learning

In this age of information overload,how to quickly and accurately filter information of interest from massive amounts of information on the Internet is a very important issue for information consumers.From the point of view of the information producer,it is also a difficult thing to let people pay attention and recognize the information they produce.The emergence and development of the recommendation system is to solve these problems and provide many conveniences for the users.The core research content of the recommendation system is the recommendation algorithm,In recent years,there are many new breakthroughs in the field of machine learning and artificial intelligence.It also promotes the development of the recommendation algorithm,and combines the new research results with the recommendation algorithm to improve the accuracy and the recommendation efficiency of the recommended algorithm,which makes the traditional recommendation algorithm have a better effect.In this paper,we combine the deep learning method with the traditional recommendation algorithm to improve the quality of the recommendation,The relevant theories and practice methods of deep learning are the key research contents of this article,the important structure of deep learning such as logic regression model,Boltzmann machine,single neuron sigmoid function and so on.We use the traditional collaborative filtering algorithm and the improved deep mixing model to carry out experiments on the standard dataset Cite ULike,to verify the performance of the new deep network structure and the deep Boltzmann machine in the recommendation scoring system.The specific research content and innovation mainly have the following three points: 1.Analyze the advantages and disadvantages of traditional recommendation systems and collaborative filtering algorithms;explain the principles and application scenarios of deep belief networks and common depth models.2.Combining an ordinary Boltzmann machine with a softmax model to form a new deep confidence network,conducting experiments on an open source data set,verifying its excellent feature extraction capabilities,and demonstrating the rationality of the new model.3.The new proposed depth model is applied to collaborative filtering recommendation algorithm,and the new collaborative filtering algorithm based on depth learning is evaluated using the evaluation index of the recommendation algorithm.The results are verified by the Cite ULike data set,and the traditional collaborative filtering algorithm is compared.It is proved that the proposed algorithm has better accuracy and efficiency.