Research On Recommendation Algorithm Based On Deep Learning

With the rapid development of information technology,people's concern is how to obtain effective information from massive data.For example,the e-commerce industry provides a variety of options so that users cannot quickly obtain information.Therefore,a recommendation system has emerged,which brings convenience to both the platform and users.The traditional recommendation algorithm mainly uses the user's historical behavior records and interests and item characteristic data to analyze the user's potential characteristic preferences and recommend the analyzed recommendation list to the user.However,traditional recommendation algorithms suffer from data sparseness and cold-start problems,which lead to inaccurate recommendation results.In order to further improve the performance of recommendations,the introduction of deep learning into recommendation systems has become a research hotspot.Deep neural network models in deep learning can effectively obtain feature representations,thereby reducing the sparseness of the data.Compared with traditional collaborative filtering algorithms,the application of deep learning makes the recommendation effect more accurate.In response to the problem of sparseness of the original data,this paper proposes two improved algorithms.The first algorithm is a Neural Cooperative Filtering Algorithm(NSSCF)combining Singular Value Decomposition(SVD)and Stacked Denoising Auto-encoders(SDAE).The algorithm first uses SVD to reduce noise;then uses SDAE's effective feature learning to obtain item features and calculate the similarity between items based on the score;then calculates the similarity between items based on the attributes on the attribute item matrix;finally calculates the comprehensiveness of the items Similarity and the collection of nearest neighbors between items,thereby improving the accuracy of the algorithm.In order to learn the potential characteristics of users and projects more effectively,a collaborative filtering algorithm based on deep learning feature representation(DLFeaCF)is proposed based on explicit feedback and implicit feedback.This algorithm is mainly based on inner product and outer product.It uses Multi-Layer Perception(MLP)and Convolutional Neural Network(CNN)to learn the global and local features of user-items.Finally,the features are combined at the fusion layer and a prediction score is obtainedFinally,experiments were performed on the real MovieLens data set and compared with other algorithms.It was shown that the proposed algorithm improves the accuracy and effectiveness of the recommendation algorithm to a certain extent.Figure[24]Table[7]References[62]...