期刊名称：International Journal of Combinatorial Optimization Problems and Informatics
印刷版ISSN：2007-1558
电子版ISSN：2007-1558
出版年度：2015
卷号：6
期号：2
页码：11-27
语种：English
出版社：International Journal of Combinatorial Optimization Problems and Informatics
其他摘要：We p resent a novel ap p roach to find relevant features for identify ing four subty p es of Guillain - Barré Sy ndrome ( GBS ) . Our method consists of a combination of Quenching Simulated Annealing (QSA) and Partitions Aro und M edoids (PAM ), named QSA - PAM method. A 156 - feature real dataset containing clinical, serological and nerve conduction test data from GBS p atients was used for exp eriments. Different feature subsets were randomly selecte d from the dataset using QSA . New datasets created using th ese feature subsets wer e used as inp ut for PAM to build four clusters, corresp onding to a sp ecific GBS subty p e each. Finally , p urity of clusters was measured. Sixteen features from the original dataset were encountered relevant fo r identify ing GBS s ubty p es with a p urity of 0.8992. This work rep resents the first effort to find relevant features for identify ing GBS subty p es using comp utational techniques. The results of this work may help sp ecialists to broaden the understanding of t he differences among subty p es of GBS .
其他关键词：feature selection for clustering; search op timization; hy brid methods for clustering.
