The Fraud Detection in Individual Medical Crowdfunding Based on a Bipartite Graph

The rapid development of social media has profoundly influenced the cause of Individual Medical Crowdfunding (IMC), driving the success of the campaign. However, with the explosion of fraudulent news, there is a crisis of trust in IMC platforms. Previous research on IMC fraud detection is based on text features, this paper identifies potential frauds from anomalous social relationships. We use an unsupervised fraud detection scheme and improved FRAUDAR for edge suspiciousness constructed on the campaign-endorser bipartite graph. Based on the real data from Shuidichou, we find that the number of endorsers has a significant relationship with fundraising, but not much correlation with campaign authenticity. Our model can effectively identify numerous campaigns that are difficult to observe, and maintain a high recall. In this case, the model we put forward can be used as an ancillary tool for IMC platforms to detect potential risks in anti-fraud efforts