Research On New Retail Enterprise Commodity Intelligent Recommendation Mechanism And Application Based On Neural Network

In the retail field,the traditional business model has been unable to meet the needs of users,so a new business model-e-commerce has emerged,and people’s shopping habits are gradually changing to the Internet.However,with the rapid development of e-commerce,various kinds of information emerge one after another,and users cannot quickly obtain the information they are interested in;For various shopping platforms,merchants cannot obtain user preferences and provide them with personalized recommendation services.Therefore,how to use recommendation system to find useful information from the massive commodity data has become an important topic.At present,Most of the current algorithms mainly make use of a series of user behaviors to make recommendations,but there are common problems such as sparse matrix and low recommendation accuracy.In view of these problems,this paper puts forward the following improvements:(1)To solve the problems of matrix sparsity and cold start in traditional algorithms,this paper studies and proposes a collaborative filtering recommendation algorithm integrating product features.First,using variance analysis method,mutual information method to select the product features.Secondly,using Word2 vec model to construct product feature vector and calculate and the similarity matrix based on product feature.Finally,this paper improves the traditional collaborative filtering algorithm,which is a linear combination of the similarity matrix based on product features and the collaborative filtering similarity matrix.Experimental results show that the improved algorithm improves the accuracy and recall rate of recommendation results to a certain extent.(2)In order to solve the problem that the traditional recommendation algorithm has insufficient use of auxiliary information,this paper studies and proposes a feature fusion recommendation model based on neural network.First,using one-hot encoding to construct feature vectors.Secondly,the neural network embedding layer is used to obtain low-dimensional and dense feature representation;Finally,this paper constructs a feature fusion recommendation model,which uses the product of the user feature matrix and the item feature matrix to represent the user’s predicted rating for the item.Experimental results show that the proposed feature fusion recommendation model based on neural network improves the mean square error,root mean square error and mean absolute error.(3)This paper designs and implements the product intelligent recommendation system of new retail enterprises to provide intelligent recommendation service for users.To sum up,this paper mainly studies the recommendation algorithm based on the combination of commodity features and neural network.Then a recommendation system based on the above algorithm is designed to provide personalized recommendation service for users and improve user satisfaction.