A fast feature selection algorithm applied to automatic faults diagnosis of rotating machinery

This work presents a fast algorithm to reduce the number of features of a classification system increasing the performance without loss of quality. The experiments show that the proposed algorithm can reduce the number of features quickly as well as increase the quality of the predictions simultaneously. Three features extractions were used to generate the initial pool of features of the system. Comparative results of the proposed algorithm with the classical sequential forward selection algorithm are shown.Keywords: feature selection, feature extraction, fault diagnosis, rotating machinery, supervised learning