Research And Application Of Deep Collaborative Filtering Algorithm In Recommendation System

The rapid development of the Internet has accumulated a large amount of data and product information,resulting in the problem of information overload.Personalized recommendation has increasingly become a key solution to overcome information overload and has been adopted by many news media and e-commerce websites.In the current personalized recommendation method,model-based collaborative filtering algorithm is widely adopted due to its advantages of simple engineering implementation and strong universality.However,the current collaborative filtering algorithm based on traditional machine learning model will face the problem of insufficient feature extraction,and the recommendation accuracy is often restricted.To solve this problem,this paper takes movie recommendation as an example and designs a personalized recommendation system based on deep collaborative filtering.The main work is as follows :First,this paper added additional text information contained in the film introducetion and previous audience comments in the original data set to enrich the feature space.Firstly,the text information required by the system design is obtained through Python crawler technology,and then the text is mapped to vector representation by means of Word2 Vec method,and then the vectorized text information and the attribute information of the movie itself are fused into the data set required by the system.The second one is to abandon the traditional machine learning method and adopt the convolutional neural network in the deep learning method as the feature extractor.Secondly,this paper designs a depth(ConvMF-DenseNet)collaborative filtering recommendation algorithm,this algorithm USES deep learning the convolutional neural network as a new feature extraction method,the characteristics of the first data set information after data preprocessing transformation,again through the convolutional neural network for feature extraction of deep,finally to extract the characteristics of the use of traditional collaborative filtering algorithm is recommended to users.This method utilizes the advantages of deep learning which is good at data feature extraction and collaborative filtering algorithm which is good at similarity prediction score.The algorithm model integrated with deep learning and collaborative filtering algorithm reflects the advantages of the above two aspects,which significantly improves the recommendation accuracy of the designed personalized recommendation system.Thirdly,this paper is based on the browser and server architecture to build personalized recommendation system interface,the user is accurate real-time personalized recommendation service,it is easy to collect information on the history of user behavior and improve the user experience satisfaction,the experiments show that compared with the existing classic recommended system,design of the system further shows the deep learning in recommendation system application prospect and development potential.