A loan application fraud detection method based on knowledge graph and neural network
Qing Zhan, Hang Yin · 2018
Dianrong, a tech-driven internet finance company, provides loans to a large number of people and small business users. The ability to predict fraud from loan applications is key to the company's business. Based on published literature on fraud detection techniques, features have to be extracted manually for further rule design or machine learning. But as fraudulent behaviors change over time to avoid detection, simple features or rules become obsolete quickly. Normally, we have to extract hundreds of features, which is a time and resource consuming process. This paper proposes a new way to extract features automatically from a borrower's phone network graph using neural networks, which not only overcomes the above issue, but also captures features that are hard to fake. This method has yielded strong results in reality.