Credit Risk Prediction using Ensemble Machine Learning Algorithms

Abstract

Credit risk is a significant focus in the banking and finance industry since evaluating the borrower's ability to repay a loan is crucial before extending credit. Also, in emerging nations, the underbanked population lacks access to the collateral and identification often necessary by banks before they will issue loans. This research study proposes a novel approach for predicting credit risk in financial institutions using ensemble machine learning models. The data is preprocessed, and relevant features are selected by evaluating the feature's importance using the information gain method. The first ten relevant features are selected for training the machine learning models. To predict credit risk, the suggested method used gradient boosting algorithms, including XGBoost, XGBoost RF, and CatBoost. The proposed approach is compared with other state-of-the-art algorithms like Adaboost, Random forest, and neural networks. Moreover, the findings prove that gradient-boosting algorithms like Xgboost and CatBoost outpace other algorithms by achieving the highest training accuracy of 93.7% and 93.6%, respectively, and testing accuracy of 93.6% and 93.8%, respectively. While XGBoost takes comparatively one-third of the time for training concerning the CatBoost. Hence, the XGBoost outperforms all the models regarding the accuracy and time trade-off. Hence, the proposed approach can be applied to financial institutions to provide credit to high-security borrowers.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 41 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.