Machine learning-driven credit risk: a systemic review

Abstract Credit risk assessment is at the core of modern economies. Traditionally, it measured by statistical methods and manual auditing. Recent advances in financial artificial intelligence stemmed from a new wave machine learning (ML)-driven credit models that gained tremendous attention both industry academia. In this paper, we systematically review series major research contributions (76 papers) over past eight years using statistical, deep techniques to address problems risk. Specifically, propose novel classification methodology for ML-driven algorithms their performance ranking public datasets. We further discuss challenges including data imbalance, dataset inconsistency, model transparency, inadequate utilization models. The results our show that: 1) most outperform classic estimation, 2) ensemble provide higher accuracy compared with single Finally, present summary tables terms datasets proposed