New Generation Computing, 21(2003)135-147
Ohmsha, Ltd. and Springer-Verlag
Received 28 February 2002
One of the most important characteristics of chance discovery is that it focuses on the specific events or patterns in which the essential nature of an applied domain is implicitly included. The understanding and forecasting of such patterns and events will have a significant impact on decision making in the applied domain. This paper discusses the meaning of chance discovery from the viewpoint of medicine. Since chance discovery in medicine can be viewed as the way to find a suitable occasion for some critical actions or to check the dangerous possibilities, called rare risky events, detection and interpretation of rare but important events are ones of the components that supports chance discovery. According to this observation, several approaches for detecting rare events were introduced and evaluated by a small dataset on neurological diseases. Experimental results show that a set of events which include rare risky events can be detected by the introduced detection method, though interpretation by domain experts is required for selection of such events.
Keywords: Time-series Analysis, Fuzzy Qualitative Trend, Rough Sets, Rule Induction, Rare Risky Events.