Applying data mining to detect fraud behavior in customs declaration

This paper introduces a data mining approach to detect fraud behaviors in customs declaration data. Some of the data mining technologies used in this project, such as an easy-to-expand multidimensional criterion data model and a hybrid fraud-detection strategy, are considered. Due to the characteristics of the data distribution in fraud detection applications, it is more difficult to predict the fraud behaviors. However, the easy-to-expand data model with multidimensional-criterion introduced in this paper improves both the accuracy of the model and performance of the algorithm. Since this model has a strong ability of popularization, it can be used as a reference to other similar complex applications.