Journal
JOURNAL OF THE FRANKLIN INSTITUTE-ENGINEERING AND APPLIED MATHEMATICS
Volume 356, Issue 8, Pages 4716-4731Publisher
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.jfranklin.2019.01.046
Keywords
-
Ask authors/readers for more resources
Since the risk of loan defaulting in peer-to-peer (P2P) lending is notoriously difficult to evaluate, a deep neural network-based decision-making approach is proposed in this work for more effective assessment of P2P lending risks. Although normally a dozen features were used for neural network modeling in previous studies carried out by other researchers on similar topics, more comprehensive features including both numeric and categorical ones (e.g. home ownership and purpose of loan), are considered in this work for improved modeling. Since categorical data cannot be used directly as the input of neural networks, they are converted to numerical data using one-hot encoding function. The deep neural network (DNN) used in this work is a multilayer perceptron (MLP) with three hidden layers trained by the back-propagation algorithm. In empirical analysis, the loan data issued by the Lending Club through 2007-2015 are classified into three classes, i.e. safe loan, risky loan and bad loan using TensorFlow. The training and test data sets consist of 221,712 and 55,428 data observations, respectively. Since most of the data belong to the class of safe loan, Synthetic Minority Over-Sampling Technique (SMOTE) is used to improve the DNN prediction accuracy. It is shown that with the proposed approach the test data are classified at an accuracy of 93%, which is much higher than the predication accuracy of 75% obtained using MLP with only one hidden layer. (C) 2019 Published by Elsevier Ltd on behalf of The Franklin Institute.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available