Research On Hybrid Recommendation Algorithm Based On Deep Learning

Recommendation algorithm is one of the important technologies in Internet applications.Traditional recommendation technologies include collaborative filtering,factorization machine and so on.However,with the advent of the era of big data,the number of Internet users and items is increasing,the choices of users are more and more diverse,and the interest migration occurs more and more frequently,which makes the system to capture user interest more difficult.The corresponding disadvantages of the traditional recommendation technology have also been exposed.The recommendation method based on user similarity or item similarity can not meet the changing recommendation needs of users gradually,and the recommendation quality of the traditional recommendation system is also declining rapidly.As a method to solve complex problems in machine learning,deep learning has been successfully applied to personalized recommendation systems.The recommendation algorithm based on deep learning has achieved many results through the application of deep learning in the field of recommendation,and it has realized the characterization of the diversity of user interests in the modeling of user interests,so as to recommend the content of interest for users more accurately,solve the many drawbacks of traditional recommendation algorithms,and improve the utilization of information and the accuracy of recommendation.The existing recommendation algorithms often ignore the comprehensive consideration of users’ long-term and short-term interests,and only consider the interests from a single perspective of long-term or short-term,which leads to the problem of low recommendation efficiency and affects the efficiency of the algorithm.This paper focuses on the limitations of these algorithms,and proposes two improved algorithm models,which integrate the long-term and short-term interest preferences of users,and complete the new item recommendation for users.The specific work content is as follows:(1)By integrating the user’s long short-term interest,the user’s personal interest can be more accurately depicted,and the problem of reducing the recommendation effect caused by the dynamic change of user interest can be avoided.Therefore,this paper proposes a Recommendation Algorithm based on Long Short-Term(RA＿LST).The algorithm uses the latent factor model and the gated recurrent unit to capture the user’s long short-term interest respectively,and then uses the stochastic gradient descent optimization algorithm to fuse the two interest simulation results to output the best recommended item ranking for the user to realize the prediction of the user’s interest.(2)At the same time,considering the influence of the weight allocation accuracy of long short-term interests on the recommendation results,on the basis of the previous improved algorithm,the attention mechanism is further introduced into the model.This paper proposes an Attention Mechanical Recommendation Algorithm based on Long Short-Term(AM＿LST).The algorithm uses the attention mechanism to assign weights to long short-term interests by learning the internal relationship between interest features,which makes the fusion of interests more flexible and closer to the user’s real interests,so as to make the recommendation results more accurate.The experimental results on Movie Lens and Netflix datasets show that the improved algorithm proposed in this paper can effectively balance the influence of user’s long short-term interests on the recommendation results,and avoid the disadvantages caused by the single consideration of user interests.Compared with the algorithm before improvement,the algorithm proposed in this paper can predict the user’s interest more effectively,the recommendation accuracy has been significantly improved,and the recommendation quality of the recommendation system is optimized.