Credit Risk Prediction using Genetic Algorithm and Stacking Ensemble

It is not an easy task to estimate the likelihood that a borrower will fulfill his or her duty to the lender (bank). There are several factors to take into account before deciding whether to extend any kind of loan. This decision-making process calls for sufficient quantitative data to compute, which takes time and is susceptible to error by humans. To finish this difficult and busy work, An Machine learning model for Credit Risk Prediction using Stacking Ensemble (ESCRPGSE) has been proposed as a solution to this time-consuming and difficult task. It combines oversampling technique, ‘Gaussian Smote’ to solve the imbalance problem, Genetic Algorithm to select useful features that are useful for model training, and a Stacking ensemble bagging technique for classification. By implementing 10-fold cross-validation, the proposed model’s performance is evaluated in terms of accuracy and f1-score, and it is then contrasted with standard single-classifier classifier models, ensemble-based models, and various models that have been published in the literature.