Guardians of Credit: "A Comprehensive Analysis of Machine Learning Models for Credit Risk Detection"

Credit cards are widely used in today's digital and cashless world, however they operate at the risk of borrower to repay their credit loan, often times causing huge losses to the financial sector. Identifying such risks and preventing them beforehand is crucial. This paper presents a comprehensive study for credit risk detection by making use of different machine learning algorithms to classify customers into 'good' and 'bad' credit categories. This paper helps identify such risks and hence prevent them beforehand. The ANN model outperforms others in terms of the accuracy and ROC curve with an accuracy of 85.88%. Random Forest model provides the most balanced confusion matrix followed by confusion matrix of XG-Boost. Synthetic data points were generated to balance the imbalanced dataset, and feature engineering techniques is used to eliminate less relevant features. The study aims to help financial institutions reduce credit loss risk and identify creditworthy customers.