Spatial-temporal Attention Graph Neural Network for Fraud Detection

PROJECT TITLE :

Graph Neural Network for Fraud Detection via Spatial-temporal Attention

ABSTRACT:

Card fraud is a significant problem that results in significant financial losses for cardholders as well as the banks that issue cards. The most up-to-date techniques make use of approaches based on Machine Learning in order to identify fraudulent behavior from transaction records. But because manually generating features requires domain knowledge and may lay behind the method of operation of fraud, we need to automatically focus on the most relevant fraudulent behavior patterns in the online detection system. This means that we need to implement a pattern recognition algorithm. As a result of this, the work that we are presenting here proposes the use of a spatial-temporal attention-based graph network (STAGN) for the detection of credit card fraud. In particular, we begin by training a graph neural network to understand the characteristics of the transaction graph based on both time and location. Following that, we apply the spatial-temporal attention on top of the learned tensor representations, which are subsequently fed into a three-dimensional convolution network. Using 3D convolution and detection networks, the attentional weights are jointly learned in an end-to-end fashion with the help of joint learning. After that, we move on to the real-world card transaction dataset and run a series of extensive experiments on it. The conclusion drawn from this finding is that STAGN outperforms other state-of-the-art baselines in terms of both its AUC and its precision-recall curves. In addition, we conduct empirical studies with domain experts on the proposed method for fraud detection and knowledge discovery. The results demonstrate the superiority of the proposed method in detecting suspicious transactions, mining spatial and temporal fraud hotspots, and uncovering fraud patterns. It is demonstrated that the proposed method is effective in performing other tasks that are based on user behavior. In conclusion, in order to meet the challenges posed by Big Data, we integrate the STAGN that we have proposed into the fraud detection system in order to use it as a predictive model, and we present the implementation detail of each module contained within the system.

Did you like this research project?

To get this research project Guidelines, Training and Code... Click Here