CredShield: Decentralized AI for Secure and Adaptive Credit Risk Management

The amount of credit card default prediction is large because banks and financial institutions are searching for a method to predict cards that can be chosen according to the borrower’s credibility. This work predicts how people will pay for credit with transactional and demographic data and runs a few models of machine learning, i.e. logistic regression, random forest, and neural network. Because data can be partitioned to understand and enhance the likelihood of accurate prediction success, KMeans is combined with DBSCAN clustering to improve the accuracy of prediction as the performance of data depends on the input data, which is less than the performance of defined DBSCAN. The results of the proposed method of this UCI machine learning repository dataset were 99.32% compared to traditional models that were 82.8% and 83.5%, respectively. One is the need to do data normalization, and the other is to select appropriate methods for multifold validation to get maximum performance. Ultimately, this solid analytical system aids in finer credit risk analysis, inhibits inclusion of default, and so ends up with heart-wrenching and lucrative loan approvals.