DenseNet+Inception and Its Application for Electronic Transaction Fraud Detection
Zhuo Chen, Guanjun Liu · 2019
With the development of online payment, electronic transaction fraud events also take place often and result in huge financial losses. Recently, DenseNet as one of the most prominent Convolutional Neural Networks (CNNs) has achieved a remarkable success in many applications such as image recognition. However, when we apply DenseNet to the electronic transaction fraud detection in which the original data is directly as the input of DenseNet, a perfect performance cannot be obtained since the transaction data is different from the image data and the features of the original transaction data are not rich. The Inception module used in GoogLeNet (a kind of CNN) has different convolution kernels of different sizes and thus can enhance the feature mining ability. In this paper, we improve DenseNet (iDenseNet for short) in order to suit electronic transaction data and then propose two integration patterns of iDenseNet and Inception module. Experiments are done on two large datasets with millions of transaction records and compare our methods with CNN, DenseNet and Random Forest that are the state-of-the-art methods used in electronic transaction fraud detection. Results illustrate better performances of our methods.