Research On Intelligent Recommendation Algorithm And Recommendation System

With the rapid development of the Internet in the 21 st century,with the popularity of mobile devices,the interaction between users and the network has gradually increase,and a large amount of data is generated,which brings the problem of "information overload" to users.How to quickly and accurately push the content of daily needs to users is an important research direction of recommendation systems.Therefore,the research of information recommendation technology is valuable and meaningful.In recent years,with the rise of the wave of artificial intelligence,the fields of machine vision and intelligent devices have developed by leaps and bounds.Try to combine the recommendation system with deep learning to explore a new development direction.Aiming at the problems of traditional recommendation algorithm,such as sparse data,slow update of recommendation content,and difficulty in utilizing other features and contextual features,etc.In this paper,a recommendation algorithm DAPN based on deep learning is proposed,and combines this algorithm with traditional recommendation algorithms to design and implement a hybrid movie recommendation system.The specific research are as follows:The research background and significance of the subject are described,and the research status of deep learning and recommendation algorithms at home and abroad is analyzed;At the same time,we studied the traditional recommendation algorithm and the current main depth learning recommendation technology,analyzed their res pective advantages and disadvantages,and made improvements on the existing basis to propose new algorithms.The prediction method of movie click through rate based on neural network is studied,and the network is optimized and improved on this basis.In this paper,a deep attention module DCA is designed by improving the scaling point product attention mechanism,and a new movie click through rate prediction model(DAPN)is proposed by integrating factor decomposition(FM)algorithm and DCA module into n eural network.This model uses the improved zoom click attention module to mine the hidden interest features in user behavior,and uses the factor decomposition machine to input the neural network with the dominant preference in the user’s explicit feedbac k information to learn deeper information.Finally,the experimental results were verified and analyzed on Movie Lens-1 m and Avazu public data sets.Through the analysis of traditional recommendation algorithms,it is found that due to the constraints of da ta sparsity,"long tail" and other issues,the feature data of movies are not fully utilized,and the deep relationship between users and movies cannot be learned.This paper proposes a new hybrid recommendation model based on the advantages of many algorithms.And designed a personalized hybrid movie recommendation system that meets the accuracy,diversity and can solve the problem of sparse data.Taking Movie Lens movie data as the initial data of the system,according to the demand analysis and overall architecture of the system functions,a movie recommendation system is built using the Spark big data platform using the B/S structure.The proposed DAPN model is applied to the movie recommendation system.