Research On Hybrid Recommendation Algorithm Based On Shopping Feature Classification

With the advent of the 5G era,the industrialization and scale of the Internet Economy has ushered in a new chapter.5G technology will provide stronger technical support,more comprehensive consumption field and faster user experience for the vigorous and development of the consumer Internet era.Therefore,in the era of consumer Internet,people's life has been greatly affected and has changed greatly.People are constantly shifting the focus of consumption habits to e-commerce platform.In real life,people's life is more and more inseparable from the network,more and more closely linked,more and more inseparable from the network,so it is more and more inseparable from the network.Online shopping can not only bring great convenience to people's life,enrich people's entertainment ways and bring pleasant consumption experience,especially during the COVID-19 epidemic period,but also be one of the important economic growth models.At the same time,the websites of major e-commerce platforms constantly update and store interactive records of users' online behaviors with daily PB store level.These transaction records and other data have great value.Scientific research and utilization of valuable information will effectively improve the user's shopping guide and improve the service quality of e-commerce.Consequently recommendation system is the most popular to be adopted.In this paper,we deeply mine the user's historical behavior and continuously the data expressing the shopping intention features are classified and extracted.Then a unique algorithm is used to obtain the prediction score and we propose an enhanced rating matrix collaborative filtering recommendation algorithm(EPR?IBCF).On the premise of strong purchase intention of users,it can accurately recommend the items which they are interested into the target customers,so that users can reduce the loss of shopping guide and purchase more satisfied orders at the same time.The main research work is as follows:In this paper,firstly the user similarity matrix with potential interests is established based on the collaborative filtering recommendation algorithm.As a result of that,the basic rating matrix(BRM)will be achieved respectively.Secondly,Restricted Boltzmann Machine is used to deeply mine the characteristics of purchase intention and penalty factor is used as the evaluation weight of ERM which classify the features and weight-added will be showed as commodity profile of users' purchase intention force.Then the predictive rating score for enhanced matrix ERM could be obtained.Finally BRM&ERM will be merged with weight value to get ultimate score.Subsequently top ranked items will be recommended to the target users with stronger purchase intention based on top-N results.It effectively solves the situation that the products of interest against users rank uppermost but are not purchased or rarely purchased.The recommendation aim against more strong purchase intention and more accurate target users will be realized.The experimental results show that compared with the traditional commodity collaborative filtering IBCF recommendation algorithm,EPR?IBCF algorithm improves the recommendation precision rate by2.68%,recall rate by 4.61% and comprehensive value F1 by 3.39%.