The Design And Development Of Commodity Recommendation System Based On Improved RBM

The Internet is now showing a flourishing trend,and information overload is the feature that the Internet has given to this era.The richness and diversity of goods let consumers do not know how to choose.It is therefore recommended that the system becomes more and more indispensable,it can help the user to select the most suitable user preferences of goods from a large number of goods,in order to reduce the cost of users obtain the information of the goods,the recommendation system,how to use the information content and analysis of user behavior data and predict the user potential interest is the focus of this paper.Collaborative filtering is widely applied.Traditional collaborative filtering only uses user rating matrix,but this rating matrix has a high sparsity.It will greatly reduce the accuracy of recommendation,while there is a cold start problem for new products.Aiming at these problems,this paper improves the traditional restricted Boltzmann machine RBM.The improved algorithm GCS-RBM has better adaptability in data sparsity and cold start problems,and improves the accuracy of recommendation.The main work of this article is as follows:(1)The traditional RBM algorithm is improved,and a collaborative filtering algorithm GCS-RBM,which combines the similarity of the content of the commodity content,is proposed.This method uses Word2 vec vector of commodity content representation,and calculate the similarity between the goods,and then to the similarity measure between goods added to the RBM model to predict the score,the final prediction score not only considers the influence factor in the rating matrix,but also consider the effect of similarity between the contents of goods.(2)Using Netflix and MovieLens data sets,we use RMSE,Recall,mAP these indicators to evaluate the improved algorithm.Compared with the traditional RBM algorithm and existing classic improved algorithm,the experimental results prove that the improved algorithm GCS-RBM has better recommendation performance.(3)A commodity recommendation system based on improved RBM is excellent implemented.The recommendation system adopts B/S three tier architecture,and uses three main frameworks of Spring,SpringMVC and Mybatis,Oracle database and Nginx server to develop.The improved recommendation algorithm,GCS-RBM,is implemented through an extended interface provided by Mahout,which can efficiently manage user and commodity related information.It can recommend the goods that they are interested in.