Downloads: 1 | Views: 15 | Weekly Hits: ⮙1 | Monthly Hits: ⮙1
Research Paper | Information Technology | India | Volume 12 Issue 11, November 2023 | Popularity: 4.2 / 10
Predicting Loan Default Risk in P2P Lending Platforms: A Study of Lending Club Borrowers
Vinay Singh
Abstract: The peer-to-peer lending industry has experienced rapid growth due to increasing demand from borrowers and lenders. These platforms have done well because of digital changes and the wider reach of the Internet, which connects people of all ages and backgrounds. Lending institutions encounter significant challenges in accurately predicting loan defaults. When large loan amounts are defaulted, it results in considerable business losses. This study focuses on loan defaults in online peer-to-peer lending. The dataset "BALANCED_Data_Predicting_Default.csv" used for this research was sourced from Carmen?Ohio State University. This dataset contains 58 variables on 20, 000 actual Lending Club loans issued in 2015. This dataset was loaded to Orange3, a data mining application. The loan status was selected as the dependent variable and categorized into two groups: "default" and "fully paid" loans. The dataset was preprocessed to remove any irrelevant data. We evaluated the variance and removed variables with little variation. Some attributes were excluded based on our judgment and business knowledge. Certain columns, such as "collection_recovery_fee" and "recoveries", were considered irrelevant since they didn't provide useful insights into loan defaults. This research aims to apply AI and ML, specifically Decision Trees, logistic regression, Random Forests, SVM, Neural Networks, and gradient boosting, to predict the default probability of Lending Club borrowers. As part of this research, we will pick the best-performing Model and report its performance. If these models are used, Lending Clubs and loan companies can make data-driven decisions, enhance services, and predict customers' defaults. We have tried multiple machine learning models, including logistic regression, random forest, gradient boosting trees, support vector machine (SVM), and neural networks. We tuned the parameters of different models (e. g., the number of layers in neural networks). In this case, the gradient boosting tree performs well, as we achieved the best result, F1 0.883.
Keywords: P2P Lending, Lending Club, Orange3, Imbalanced dataset, Loan-default, Prediction, Logistic Regression, Random Forest, Gradient Boosting tree, Support Vector Machine (SVM) and Neural Networks
Edition: Volume 12 Issue 11, November 2023
Pages: 2255 - 2260
DOI: https://www.doi.org/10.21275/SR231114083515
Make Sure to Disable the Pop-Up Blocker of Web Browser