Deep Learning Model Implementation Using Convolutional Neural Network Algorithm for Default P2P Lending Prediction

Abstract

Peer-to-peer (P2P) lending is one of the innovations in the field of fintech that offers microloan services through online channels without intermediaries. P2P  lending facilitates the lending and borrowing process between borrowers and lenders, but on the other hand, there is a threat that can harm lenders, namely default.  Defaults on  P2P  lending platforms result in significant losses for lenders and pose a threat to the overall efficiency of the peer-to-peer lending system. So it is essential to have an understanding of such risk management methods. However, designing feature extractors with very complicated information about borrowers and loan products takes a lot of work. In this study, we present a deep convolutional neural network (CNN) architecture for predicting default in P2P lending, with the goal of extracting features automatically and improving performance. CNN is a deep learning technique for classifying complex information that automatically extracts discriminative features from input data using convolutional operations. The dataset used is the Lending Club dataset from P2P lending platforms in America containing 9,578 data. The results of the model performance evaluation got an accuracy of 85.43%. This study shows reasonably decent results in predicting p2p lending based on CNN. This research is expected to contribute to the development of new methods of deep learning that are more complex and effective in predicting risks on P2P lending platforms.