E-Commerce Recommendation Algorithm Based On Restricted Boltzmann Machine

With the continuous development of Internet technology,e-commerce is occupying a more and more crucial position in the development of social economy.With the gradual expansion of e-commerce market,more and more traditional industries are developing towards modern e-commerce.E-commerce provides users overmuch goods,even if it provides users with convenient shopping,it will also let users getting lost in a large number of commodity information space and unable to find what they really want.The recommendation system comes into being in this case.The E-commerce recommendation system can personalize the recommendation according to the users' preference,so that users can select the goods they need quickly and accurately in the vast amount of commodity information.In this paper,the existing recommendation algorithms is summarized and the existing problems are discussed.When the user scoring data is sparse,the similarity measure of the traditional collaborative filtering algorithm will not be accurate.To solve this problem,a collaborative filtering algorithm based on the depth network is proposed.Firstly,the stacked Boltzmann machine composed of the deep belief network is used to monitor the user's behavior without supervision.The high-dimensional and sparse user behavior is compressed into a low-dimensional and dense user characteristic vector in order to get the user's score feature.Then,the user score feature and the user project attribute preference feature are combined to get the user similarity.Finally,a recommended list is generated based on user similarity.The experiments of the Movielens dataset and an e-commerce data set show that the new algorithm can effectively improve the hit rate,recall rate and precious value of the recommended results.To a certain extent,it solves the problems the difficulty to calculate the similarity with the sparse score matrix in the project.