Research On Recommendendation Algorithms Based On Deep Learning

With the continuous development of Internet technology,the amount of data generated on the network every day has shown explosive growth,and it has become more and more difficult to mine valuable information for users from massive amounts of data.The traditional search engine has been unable to meet the personalized needs of users,and the recommendation system,as a technology that can mine the information of interest and recommend it to users based on the user's preferences in massive data,has been obtained by academia and industry Extensive research in the world.In industry,the recommendation system has played a huge commercial value in the field of e-commerce and advertising.In the academic field,many excellent recommendation algorithms have been continuously proposed to promote the continuous development of recommendation systems.However,due to the complexity of practical problems and the increasing complexity of users,the recommendation system still faces the problems of insufficient personalized recommendation,low prediction accuracy and weak stability.Therefore,this paper proposes two improved recommendation algorithm models.The work of this article is as follows:1.Under the background of deep learning,this paper combines two cutting-edge technologies in this field,namely dual autoencoder and Inception structure neural network,which are applied to the recommendation system to improve the prediction accuracy and operating efficiency of the recommendation system.Among them,the dual autoencoder(GADAE)is designed based on the gate attentionmec hanism to further improve the content-aware recommendation algorithm model of personalized products,thereby improving the personalized recommendation performance of the recommendation system.2.Further,in order to improve the prediction accuracy of the recommendation system,the NCF-i model is innovatively proposed.This model is to improve the NCF(Neural Collaborative Filtering)neural network collaborative filtering model,convolving the NCF model with the Inception structure.Combining neural networks,a neural network collaborative filtering method(NCF-i model)based on the Inception structure is proposed,and product review information is integrated into the model for prediction and recommendation.3.In order to verify the effectiveness of the NCF-i model and the GADAE model,this article is trained and tested on the public data set,and the final result is obtained: based on the dual door attention on the recommendation system public data sets The self-encoder content-aware recommendation model of the force mechanism is superior to the current mainstream recommendation model in the recall rate and normalized cumulative loss gain of Top-N recommend ation.At the same time,using the recommendation algorithm of NCF-i model,the prediction accuracy and stability of the recommendation system are better than the current mainstream recommendation model,which proves the effectiveness and wide applicability of the model designed in this paper.