Research On Cross Domain Recommendation Algorithm Based On Variational Autoencoder

Recommendation system is a system that uses machine learning,data mining and other technologies to automatically recommend relevant information,goods or services to users according to their behaviors and preferences.With the development of Internet and mobile communication,people can easily get a lot of information and goods.However,due to the large number,users need to spend a lot of time and energy searching and filtering,which makes the acquisition of information and goods more challenging.In order to improve users’ experience and satisfaction,the recommendation system comes into being,which can automatically filter out the information and products that meet users’ interests and needs.In the short video APP,the recommendation system can recommend videos in line with users’ preferences based on their viewing history,likes,comments and other behaviors,making it easier for users to find the content they are interested in.In the music APP,the recommendation system can recommend music that meets the user’s preferences according to the user’s listening history,likes,comments and other information.Traditional recommendation algorithms have problems such as high sparsity of data,inability to accurately recommend new users or new items,and lack of scalability.These problems lead to the serious cold start problem of the recommendation system.To solve these problems,the data-dense auxiliary domain can be added to expand the data-sparse data domain,so as to alleviate problems such as data sparse.Therefore,the cross-domain recommendation algorithm is introduced in this paper.Cross-domain recommendation algorithm can effectively improve the recommendation accuracy by using the potential interests of users in the source domain to extract features and apply them to the target domain.In this paper,variational autoencoder is used to extract users’ potential interests.Compared with traditional neural networks,variational autoencoder is not a black box model,and its characteristics make it easier to obtain users’ potential information,so as to improve the accuracy of the algorithm.This paper makes a detailed theoretical analysis and experimental research on the proposed recommendation algorithm,and expounds the implementation principle and feasibility of the method.In order to solve the problems of sparse data and cold start in the recommendation system,a cross-domain recommendation algorithm DAVae based on the fusion features of variational autoencoder is proposed,and the variational autoencoder is improved to achieve noise reduction effect by adding noise to it.At the same time,the influence of KL divergence weight on model performance is also studied.In order to better extract the potential features of the users in the source domain and the target domain,the cross-domain recommendation algorithm DAVae Lstm based on the VAE-LSTM fusion features is proposed based on Da Vae.As LSTM can deal with noise and incomplete data better,LSTM model is adopted as the encoder VAE to extract potential features of users in source domain and target domain.The cross-domain recommendation algorithms DAVae and DAVae Lstm based on variational autoencoder proposed in this paper are superior to the traditional recommendation algorithms in terms of recommendation performance.The experimental results show that the proposed algorithm can effectively improve the accuracy and robustness of the recommendation system,which provides a new idea and method for modern Internet application recommendation system.