Research On Product Recommendation Model Based On Similarity Division And Association Rules

With the rapid development of computer network and the development and prosperity of e-commerce in China,online shopping has gradually become a common means of shopping in modern people's daily life.Moreover,with the gradual deepening of China's the Belt and Road policy,China's trade with other countries will surely become closer.Foreign products will gradually enter the horizon of Chin's online shopping users,which are gradually accepted and favored by users.Under this background,cross-border e-commerce has developed rapidly.The number of enterprises with cross-border e-commerce business is increasing.These enterprises use the e-commerce system to provide goods to terminal consumers.The characteristic of the e-commerce system with cross-border characteristics in operation is that it is often necessary to introduce products of different origins and different brands of the same product,which makes the product categories provided in the system complicated and small in quantity.How to recommend the newly introduced products to the customer is an issue that the system must consider during the operation.Based on the actual needs of an e-commerce system with cross-border characteristics in the initial stage of operation,this dissertation conducts research on recommendation models to reduce the time for consumers to search for products and help consumers buy products that meet their wishes.The user shopping experience,while helping companies to increase the viscosity of online shopping customers,improve service quality,and then improve system operating efficiency is of great practical significance.Based on the research of related theories,this dissertation focuses on how to mine customer's need and how to implement product recommendation method.The study found that the current recommendation algorithm mainly uses collaborative filtering methods based on ratings.Because user ratings are affected by many factors,it has certain flaws in objectivity and credibility,resulting in poor recommendation effect.This dissertation proposes a recommendation algorithm based on the similarity of consumption behavior.The customer consumption orders stored in the e-commerce system are used to first divide the similarity of customers,commodities and purchase behaviors into clusters,and then use the association rule algorithm to perform mining analysis to obtain recommendation rules,and this method can better solve the problems in the collaborative filtering algorithm.In addition,in the operation of the e-commerce system,new customers are constantly emerging and new overseas products are continuously introduced into the system.How to recommend products to new customers and how to recommend new products to customers,that is,how to solve the"cold start" problem of the system In this paper,combining the clustering based on similarity division and performing association analysis on the clustering to obtain recommendation rules,a corresponding recommendation algorithm is proposed.At the same time,it also puts forward a recommendation method in daily consumption and promotion situations.The example analysis shows that the recommendation rules obtained by the association analysis after clustering by similarity have higher confidence than the association analysis without clustering,which can better improve the accuracy of the recommendation results and can solve the problem of "cold start" caused by the introduction of new users and new products has an important role in improving customer stickiness and customer satisfaction.