Abstract
Bankruptcy prediction is a very important research trend: although statistical methods are mainly used in literature, techniques based on Artificial Intelligence are interesting from many points of view. Among them, Case Based Reasoning (CBR) could be useful to cluster enterprises according to opportune similarity metrics as well as suggest proper actions to take for avoiding bankruptcy in borderline situations. In this paper, we present a new and still under development CBR approach to this problem, that seems to return better results than previous attempts. The approach is based on different kinds of similarity metrics and is focused on the implementation of innovative revise algorithms. In particular, the paper shows how the revise step is crucial to obtain better performance in bankruptcy prediction.
Keywords: Case Based Reasoning, Bankruptcy Prediction, Overall similarity
DOI: 10.1016/j.eswa.2016.07.033