Abstract :
Summary form only given, as follows. Financial Early Warning Systems (FEWS) aim to predict the probability of financial problems, such as bank failures in a dynamic and uncertain environment. They are used to generate early warnings, and provide explanations of possible reasons behind the failures. This talk will present new developments in FEWS, by applying intelligent techniques, including a novel fuzzy neural network to supply a descriptive prediction analysis in the form of fuzzy rules, a new fuzzy transfer learning framework, and fuzzy transfer learning-based prediction techniques in order to effectively use the knowledge learned from different time periods/domains with different features to support prediction and decision making in FEWS. With these developments, a EWS can use the previous knowledge to generate interpretable alerts, and support decision-makers in finding possible solutions and thus, reducing the risk of a possible problem or an event.
