Using linear discriminant analysis and data mining approaches to identify E-commerce anomaly

Abstract

Electronic commerce has been rather pervasive in our life today. However, the damage is equally pervasive. For Business to Consumer type of E-commerce, various types of E-commerce anomaly usually incurs loss of revenue, reduced customer satisfaction and compromised business confidentiality. This paper proposes linear discriminant analysis and data mining approaches to identify the E-commerce anomaly. The data mining approaches yield superior performance. However, the unbalanced data make the data mining approaches dominated by the data of the majority class. LDA is introduced to deal with the unbalanced data set. The results indicate that our proposed methods can identify the E-commerce anomaly precisely. The practice insights from the results are also given.