Research On Personalized Recommendation Model And Application Of Retail Enterprises Based On RFMQ Model

With the advent of the new retail era,it is a general trend for community retail stores to open up the online and offline shopping mode.At the same time,most of the retail stores' offline information management systems are relatively complete,collecting and storing a large amount of transaction information.This research is based on the actual demand of personalized recommendation after the transformation of online model of physical stores,and uses the mass information collected by physical retail stores to build personalized recommendation model for retail enterprises.Firstly,the shortcomings of the traditional recommendation method(PCC or Cosine)which combines mining commodity relationship and collaborative filtering are analyzed.On this basis,a novel method based on improved association rule technology to mine commodity relations is proposed.Among them,association rule method solves the limitation of using similarity method only to consider the similarity relationship of commodities in the past,and discovers the asymmetric relationship of commodities.Considering the association rules to determine the commodity relationship strategy,the first N association rules with high reliability are selected.According to the consumer purchase record and consumption information,the RFMQ model is constructed,the required fields are selected and the customer-item scoring matrix is calculated.Among them,the user's rating of items is calculated according to the weighted RFMQ model.The AHP method is used to determine the weight value,which integrates the purchase preference information of experts and users quantitatively and qualitatively.Finally,the model is modeled and merchandise relations based on association rules are integrated into the existing matrix decomposition model as regular items;the gradient descending method is used to optimize the proposed model;five-level cross-validation is used,and the rationality and validity of the method is verified on the real transaction data set Jtrade,which provides personalized recommendation scheme for supermarket enterprises.