Anti Fraud Detection Model Using Deep Learning Approach

Abstract

Recently, Internet finance has become more and more popular. However, bad debts becamea serious threat to online finance corporations. A commonly used fraud detection models by the traditional monetary companies is logistic regression. In this paper we use dataset consisting large publicloans data of a financial company i.e., Lending Club to check the potential of deep neural networks in fraud detection. Once this dataset is loaded we dealtwith the missing values and data pre-processing. With this pre-processed data, we extracted important features using the XGBoost algorithm and developed a CNN deep neural network to detect loan fraud on the Internet. Extensive experiments were conducted to prove that deep neural networks are superior tocommonly used models. This easy and effective model can give enlightenment for the utilization of deep learning to combat online loan fraud, which can profit the financial engineers of tiny and medium-sized financial corporations.