Recent techniques in data mining(including Bayesian belief networks) are able to generate very large amount of "so-called" knowledge or strings of pearls in a few minutes. However, we often observe little of them are effective in a practical sense. It is quite interesting that most strings of pearls are on a boudary between a lot of the 'nonsense', which has no meanings for anyone and small number of the 'trivial', which every one has already known. The judgement depends on the background, the objective, and the context the knowledge is used. This paper discusses these problems and investigates the performance of our data mining experience about clinical patient data provided by Prof. Tsumoto in 1999.
各種検索
サポート
ヘルプ