作者: Dominik Ślęzak , Jakub Wróblewski
DOI: 10.1007/978-3-540-48247-5_72
关键词: Pattern recognition 、 Compression (functional analysis) 、 Fuzzy set 、 Data compression 、 Set (abstract data type) 、 Statistical classification 、 Computer science 、 Rough set 、 Artificial intelligence 、 Linear combination 、 Data mining
摘要: We provide theoretical and algorithmic tools for finding new features which enable better classification of cases. Such are proposed to be searched as linear combinations continuously valued conditions. Regardless the choice algorithm itself, such an approach provides compression information concerning dependencies between conditional decision features. Presented results show that properly derived attributes, treated elements conditions’ set, may significantly improve performance well known algorithms, k-NN rough set based approaches.