Research On E-commerce Scalper Management Based On Compensate Recommendation

The rapid development of the e-commerce industry is also accompanied by some "disharmony" phenomena,such as after-sales service of product,product description and actual differences,which any customers are confused.At the same time,there are still some problems that e-commerce companies need to deal with,how to effectively identify the behavior of E-Commerce scalper(snapping up a large number of high-performance products in e-commerce platform,to get the difference profits),so that real customers get preferential treatment and improve customer satisfaction is A difficult point to be solved.Therefore,this paper analyzes and studies the problem from two aspects: on one hand,studying the identification problem of E-Commerce scalper,because the existing theoretical research is almost blank.Analyses of the purchase behavior difference between scalpers and regular customers are discussed from DFQ perspective including Diversity,Frequency and Quantity,and based upon it an effective recognition criteria system is designed.Then based on available training data,comparative studies are designed to test the recognition ability between the proposed different kinds of models.On the other hand,from the customer's point of view,the goods affected by the rushing of the cattle are personalized to the normal customers and improve their shopping.Experience and optimize the traditional LFM recommendation algorithm based on matrix decomposition to improve its recommendation accuracyHence,in this paper,the main contents are as follows:(1)In recent years,online scalper has become a serious problem impeding the proper operations of E-commerce companies.To tackle this problem,a systematical modeling research on this issue is carried out with applications to verify its feasibility in this paper.Firstly,analyses of the purchase behavior difference between scalpers and regular customers are discussed from DFQ perspective including Diversity,Frequency and Quantity,and based upon it an effective recognition criteria system is designed.Next,SMOTE algorithm is used to address the imbalance of training data.Then,two kinds of rough set theory-based algorithms including classic rough set algorithm and neighborhood rough set algorithm are utilized to establish the E-commerce scalper recognition model.Finally,based on available training data,comparative studies are designed to test the recognition ability between the proposed two kinds of algorithms.The final results show that neighborhood rough set algorithm can provide a better recognition precision,while the classic rough set algorithm can generate IF-THEN rules and easily to verify the effectiveness of DFQ criteria system.(2)Recommendation algorithms(RA)are the core of the recommendation system,and the RA based on matrix factorization is one of research hotspots.This paper focuses on the algorithm improvement of Latent Factor Model(LFM)in the matrix factorization based RA algorithms.Two basic algorithms,including batch learning algorithm and incremental learning algorithm,are modified to provide more accurate outcomes.Finally,a numerical example,which is used to carry out comparative study among different algorithms,proves that the improved algorithm is better than previous works.