Can Explainable AI Build Trust in the Financial Domain: Study of an Explainable AI Model for Bankruptcy Prediction

Can Explainable AI Build Trust in the Financial Domain: Study of an Explainable AI Model for Bankruptcy Prediction

연구 분야: Artificial Intelligence

학회: Advanced Computing and Communications Conference

The intersection of Artificial Intelligence (AI) and finance has gained increasing attention in recent years. Bankruptcy is one of the financial events that have a significant impact on the stock market and the economy. AI can be helpful in predicting bankruptcy with higher prediction accuracy than humans, which makes it very useful. While AI offers significant advantages in bankruptcy prediction and process automation, one of the main challenges is the interpretability of AI models. This leads to a lack of trust in it. In this paper, we are trying to solve this problem using popular explainable AI Algorithms: LIME and SHAP. It will help to identify features that are important in predicting bankruptcy by the AI algorithm. Experts in the stock market and the economy can test and verify the features and gradually create trust in them. To make the study comprehensive, we used four different AI algorithms and two explainable AI algorithms on a standard bankruptcy UCI dataset.