Research And Implementation Of Deep Learning Recommendation Algorithm Based On Denoising Autoencoder

In the new era of information explosion,a large amount of information has caused tremendous trouble for people to screen effective information.In order to help people quickly and effectively filter out the required information,a personalized recommendation system came into being.As the core of the recommendation system,the recommendation algorithm has always been the focus of research.Among many recommendation algorithms,collaborative filtering algorithm is the most widely used.Faced with the continuous increase of users and project data of various recommendation systems,data sparsity and recommendation efficiency have gradually become the main factors restricting the development of collaborative filtering recommendation algorithms.In recent years,with the development and application of deep neural network technology,deep learning technology can solve the bottleneck problem of collaborative filtering algorithms.Therefore,the recommendation algorithm based on deep learning method has become a hot spot in the research recommendation algorithm field.This paper studies the problem of data sparsity in collaborative filtering recommendation algorithm,improves the self-encoder recommendation algorithm,improves the RMSE effect by the proposed DeepAutoRec algorithm,and increases the RMSE effect by 2.4%.The Gaussian noise is added to the DeepAutoRec and the experimental result is higher than the U-AutoRec recommendation algorithm.Increased by 4.4%,2% more than DeepAutoRec's results.The main work of this article is reflected in the following two aspects:First,in the recommendation algorithm,because the scoring matrix is extremely sparse,it greatly affects the recommendation effect of the recommendation algorithm.Based on the self-encoder recommendation algorithm,this paper deepens the network depth of the self-encoder based on the self-encoder recommendation algorithm,and then applies a new activation function to adjust Dropout and other parameters,and proposes a new algorithm: named For the DeepAutoRec algorithm.Experimental results show that the new algorithm can improve the accuracy of the prediction score.Second,due to the inherent noise removal characteristics of the encoder,this paper proposes a deep learning recommendation algorithm for the denoising self-encoder: In the input layer of the DeepAutoRec recommendation algorithm,noise is introduced in two ways: The first noise is A certain proportion of the input data is set to zero(Mask noise);the second type of noise is a score matrix for the entire data set,and Gaussian distribution noise is added to the input layer.Through experimental verification,the algorithm can improve the accuracy of the prediction score,and the second noise selfencoder recommendation algorithm is better.