Product Recommendation Methods For Specific Types Of Users

With the development of information technology and e-commerce,the number of users' transactions and the number of commodities are increasing.Business big data has brought valuable opportunities and severe challenges to the industry and academia.While people are more exposed to rich commodity information,the cost of finding suitable products is getting higher and higher,and the problem of information overload has become more and more serious.Commodity recommendation is an important way to solve this problem.Therefore,how to recommend products to users from a mass of commodities is one of the research hotspots in recent years.Through product recommendation,on the one hand,users' shopping experience and service satisfaction can be improved.On the other hand,they can also help merchants to obtain more sales and sales profits.At present,commodity recommendations still face challenges such as data sparseness and long tail problems,which restrict the development of business recommendation systems seriously.In order to solve the above two problems,based on the real data set and the actual demand,this paper studies the product recommendation methods for specific types of users based on the diversity of product categories and user categories in the actual business environment.The main contributions of this article are:1.Using the idea of classification algorithm,a method for intelligently identifying specific users under the business data environment is designed.Firstly,the method selects the seed commodities,then uses the strong association rules of the seed commodities and other commodities,assigns the commodity weights hierarchically,then maps the weights to the users to obtain an eigenvalue.Finally,identifing the user types according to the user eigenvalues.This method can perform well that uses only shopping record data.The experiments are taken as examples to demonstrate the effectiveness of this method which on real data sets and the identification of home users of infants and young children in supermarkets.2.Based on the idea of collaborative filtering algorithm,a product recommendation method is designed.The method establishes an implicit semantic model to complete the recommendation.By decomposing the eigenvector matrix into two eigenvector matrices which are much smaller than the original matrix,the loss function is established,and the recommendation problem is converted into a problem of finding the optimal solution.It uses the principle of the least squares method to gradually fit to obtain unknown eigenvalues.Finally,relying on the data of the nearest neighbors,complements and predicts a large number of missing feature values of users,and performs better when the data is sparse and the scale is huge.At the same time,this method has improved the method by reducing the weight of penalty factors.The control of the eigenvectors of hot commodities will not explode in the process of algorithm iteration,effectively reducing the impact of popular products on the recommendation results.3.This paper builds a distributed computing platform that can deal with massive data,and carries out specific user identification and product recommendation for supermarket users,which has strong commercial value.