Tytuł pozycji:
Evolutionary algorithm to derive classification rules from sets of example
19 stron ; 21 cm
19 pages ; 21 cm
The article proposes a method to derive classification rules that correctly describe all the examples belonging to a class and do not describe all the examples not belonging to this class. The method bases on an evolutionary algorithm with dedicated to that problem specialized operators and a method of valuing their behavior. The new concept of the proposed method is that every solution obtained from the algorithm (every member of the population in the evolutionary algorithm) contains rules which describe all classes of the training data. So it is a complete solution that covers all the examples presented to the algorithm.
Bibliografia s. 18
Bibliography p. 18