Research On E-commerce Fraud Indentification Model Based On Data Mining

With the rapid development of the Internet,e-commerce has become a mainstream business model.It brings to the enterprise great opportunities,it also brings risks and losses.Because of the virtual and hidden features of network,e-commerce fraud transactions increasingly rampant,and seriously hampered the development of electronic commerce.Therefore,effectively detect fraud has become particularly important for the healthy development of e-commerce.Based on the research of the prevalent fraud transactions in e-commerce,this paper proposes to use data mining technology to build e-commerce fraud detection model.Firstly,this thesis briefly introduces some knowledge of e-commerce fraud and other industry's research about fraud detection,and then it propose to use data mining classification algorithm to build our models.Furthermore it deeply studies models of,data mining,focusing on support vector machine and decision tree.Considering unbalanced distribution of sets,low-efficient single classifier,and the cost of misclassification,it puts forward k-means clustering algorithm for unbalanced distribution of sets and combined classifier based on support vector machine and decision tree,Adacost algorithm for classifier effectly.Finally,we use voting method to combine the result of multiple classifiers,and use Coverage Rate,Rccuracy Rate and F-measure of minority class to evaluate the results of the classification of minority class sample..This paper uses the historical transactions data of an e-commerce business,and build realistic models on Weka platform.During the experiment,we build single classifier with undivided samples,single classifier with divided samples and combination classifier with divided samples,and analyze the results of each classification model to validate the correct use of k-means clustering,Adacost algorithm and combination classifier.Therefore,it proves that the models built in this paper is fit for e-commerce fraud detection.