A comparative study on feature selection to design reliable fault detection systems

A fault detection system, which is regarded as a classification task, is developed in this work. The classes considered here are the nominal or malfunctioning state. To develop a decision system it is important to select from the whole data set collected by the supervision system, only those carrying relevant information related to the decision task. A comparative study of five feature selection algorithms applied to the fault detection data is presented. The algorithms are tested on real medical diagnosis and fault detection benchmarks. The results obtained indicate that a small number of measures can accomplish and perform the classification task. The proposed classification method is compared with three well-known classifiers. © 2012 Praise Worthy Prize S.r.l. - All rights reserved.