Research On The Recommendation Algorithm Based On Distributed Representation And Recursive Neural Network

With the development and maturity of Internet,Internet of things,e-commerce,artificial intelligence,cloud computing,mobile computing,etc,there will be lots of information,products and resources,which will lead to the advent of big data era.In this age,Information overload has become one of the core issues that need to be solved in every field,The technology related to the recommendation system has become an effective way to solve this problem.In recent years,Deep learning techniques have achieved a lot in image processing,speech recognition and natural language processing.In addition,deep learning technology is also applied in the field of recommendation system and information retrieval,however,the corresponding research results are not many,so this paper applies deep learning techniques to the field of recommendation system.This paper makes an in-depth understanding and research on the concept,existing problems and current popular recommendation algorithm of the recommendation system,some common problems include cold start of recommendation system,the advantages and disadvantages of each algorithm are compared.After studying and analyzing the current status of recommendation system,the following research is carried out:(1)In view of the implicit feedback behavior in e-commerce websites,this paper selects the next shopping basket recommendation in the field of e-commerce as the research objective.In this paper,the user behavior is first processed into the time series form,a recursive neural network model of deep learning is used to extract characteristics of users' purchasing behaviors.(2)In view of the high complexity of the recursive neural network,we draw on the distributed expression method used in natural language processing.we use the open source tool of word2 vecr to learn the distributed expression of object word vectors and convert the original one-hot encoding of the object word vector into a continuous word vector.This vectoris then used as input to the recursive neural network model.On the one hand,it reduces the complexity of calculation and improves the accuracy of recommendation system.(3)The experimental results on the e-commerce site real data set and the MovieLens dataset show that,compared with the traditional recommendation method,the proposed algorithm framework has strong feature extraction capability and generalization ability,which improves the accuracy of recommendation system.