Loan-Busters is a data-driven project aimed at predicting loan defaulters using advanced data mining techniques. This project leverages a rich dataset from the Lending Club and implements Decision Tree and Logistic Regression models to evaluate the likelihood of loan repayment failures.
Data Analysis: Comprehensive exploratory data analysis (EDA) including correlation, handling imbalanced data, and categorical feature analysis.
Model Implementation: Detailed implementation of Decision Tree and Logistic Regression models.
Performance Evaluation: In-depth model comparison and performance evaluation using ROC curves, precision, and recall metrics.
Data Source: Utilizes data from LendingClub.com (2007-2010) with 9,578 records.
High-Risk Indicators Identified: Analysis revealed specific borrower attributes, such as credit score and debt-to-income ratio, as significant indicators of loan default risk.
Model Performance: The Decision Tree model demonstrated high precision in identifying defaulters, while the Logistic Regression model excelled in overall accuracy.
Imbalanced Data Handling: Techniques like SMOTE (Synthetic Minority Over-sampling Technique) were effectively employed to address data imbalance, improving model reliability.
Predictive Insights: The project provided valuable insights for financial institutions in refining their loan approval processes, potentially reducing default rates.