Research On Hybrid Recommendation Algorithm Based On Denoising Autoencoder

In recent years,the amount of online data has grown exponentially.The main part of this data is related to internet-based e-commerce platforms and social platforms.For individuals or organizations,it is very difficult to extract the required information in time from huge amounts of data.The recommendation system provides an automated and efficient solution to this problem.However,whether the information is explicitly expressed or implicitly feedback by the user is extremely sparse,which will severely limit the performance of the recommendation system.In recent years,some recommendation schemes incorporating social information have appeared,but they are also limited by the sparseness of social information and cannot further improve the recommendation performance.Because deep learning technology has the ability to mine deep features from a wide range of data,it has attracted the attention of more and more scholars.Based on the denoising autoencoder in deep learning,this article completes the project recommendation by incorporating social information.The main research contents of this article are as follows:1.A social recommendation algorithm based on denoising autoencoder is proposed.Aiming at the problem of inaccurate prediction caused by data sparsity in scoring prediction scenarios,the denoising autoencoder is used to integrate user rating information and trust user rating information in a framework by feature fusion to achieve hybrid recommendation design,and the impact of scoring data sparsity is reduced by modeling the potential impact of social trust user preference information.On this basis,the impact of trusted user scoring preferences on different types of users is distinguished through user clustering and personalized weights.Finally,experimental simulations are performed on open source datasets.The results show that the proposed method can effectively use social information to improve the accuracy of score prediction.2.An implicit feedback recommendation algorithm based on denoising autoencoder is proposed.In the absence of explicit user rating data,by fully tapping the potential of social information,the impact of sparseness of implicit feedback data is reduced and the quality of recommendations for cold-start users is improved.First,from the perspective of improving the sparseness of social trust information,a new measurement method of trust similarity is used to mine the implicit trust relationship from the overall user trust matrix and calculate a more accurate user trust value.Then use the dnoising autoencoder to deeply integrate the user's implicit feedback data and social trust data in a way to share user preference characteristics and social trust characteristics in the middle layer,and improve the quality of implicit feedback recommendation by integrating the effects of the two.