Maximizing the lender’s profit: profit-oriented loan default prediction based on a weighting model

Loan default risk prediction is necessary in credit risk assessment, as it helps financing institutions and investors make decisions. However, existing prediction models focus more on using individual classifiers to obtain higher prediction accuracy, which is far from the core purpose of business (i.e., maximizing profit) and leaves opportunities to explore profit-oriented and interpretable weighting models. This study proposes a profit-oriented weighting model for loan default prediction. The model consists of three stages: constructing multiple profit-oriented sub-classifiers, determining profit-oriented weight coefficients, and providing interpretable analysis. Five lending datasets are examined based on accuracy and profit-based metrics. The empirical results demonstrate that the proposed weighting prediction system helps lenders achieve higher profits and provides concise and intuitive interpretability. Thus, it can help practitioners make better decisions and manage risk.