Special issue on chance discovery

Abstract

More than five years have passed since Ohsawa proposed the concept of chance discovery [2]. When Ohsawa proposed the concept, data mining was already in boom and many researchers were starting, or absorbed in, studies on data mining. This trend seemed to be reasonable. The real lives of human are complex, and the future is not predictable. In order to have better or the best benefits, it is necessary to predict the future trends. Actually, in the usual case, data mining techniques provide us with satisfactory results enough for doing good business. For example, the phenomenon that subprime loans incited the market panic in August, 2007 was predictable and the risk has been pointed out by a large number of economists, due to their knowledge about many similar cases and about explicit signs of the market panic. Such a (not very but sufficiently) frequent pattern can be obtained by data mining, as far as economists have data on the events preceding the panic. On the other hand, we encounter exceptional events where simple data mining techniques and statistical analysis can not deal with. For example, the economic “bubbles” were found to die out suddenly after several years of prosperity – as did in Japan in 1991. Such an event is hard to predict, or even to explain the causality after its occurrence, because its causal relations with other events are unknown. In the latter case, i.e, if we cannot to predict the risk, the result may be the start of an even more serious scenario than the panic itself. For the Japanese case, since we could not predict the end of the economic bubble phenomenon, we suffered from serious influences of the break of the bubble economy for more than ten years. According to the posterior analysis, there are implicit (not noticed due to the rarity or the novelty) events which can be signs for the fatal, or sometimes for an extremely beneficial, scenario. Because these signs are novel, and are hard to be related to the result, it has been hard to catch them for making a suitable decision at a suitable time. And, had we a suitable decision at the time of significant signs events, we might have escaped from the hard scenario after 1991. Thus for us, it is important to determine implicit symptoms to risks or benefits (opportunities). Accordingly, Ohsawa proposed chance discovery in 2000. The definition of chance discovery is as below;