Author/Editor     Gamberger, Dragan; Lavrač, Nada; Grošelj, Ciril
Title     Uporaba izločitvenih pravil za zanesljivo diagnostiko
Translated title     Exploiting exclusion rules for reliable diagnosis
Type     članek
Source     Inform Med Slov Print Ed
Vol. and No.     Letnik 5, št. 1-2
Publication year     1998
Volume     str. 23-9
Language     slo
Abstract     Medicine and other critical applications require the use of reliable diagnostic rules, since a single wrong prediction may be very dangerous or costly. The problem of the induction of reliable rules is hard because theoretically no induction process by itself can guarantee the correctness of induced hypotheses. Additionally, in practice the problem is even harder due to the presence of noise in training instances and due to training instances that may not be representative for all the domain space. This paper presents a method that allows for the construction of reliable hypotheses from the available data. The induced hypotheses in the form of exclusion rules enable trustworthy predictions and provide the information about the prediction confidence level.
Descriptors     DIAGNOSIS, COMPUTER-ASSISTED
CORONARY DISEASE
DECISION MAKING, COMPUTER-ASSISTED