A hybrid classification approach to ultrasonic shaft signals
Lee, Kyungmi, and Estivill-Castro, Vladimir (2005) A hybrid classification approach to ultrasonic shaft signals. In: Proceedings of 17th Australian Joint Conference on Artificial Intelligence (3339), pp. 284-295. From: 17th Australian Joint Conference on Artificial Intelligence, 4 - 6 December 2004, Cairns, QLD, Australia.
PDF (Published Version)
Restricted to Repository staff only
In many applications of machine learning a series of feature extraction approaches and a series of classifiers are explored in order to obtain a system with the highest accuracy possible. In the application we discussed here at least two feature extraction approaches have been explored (Fast Fourier Transform) and Discrete Wavelet Transform, and at least two approaches to build classifiers have also been explored (Artificial Neural Networks and Support Vector Machines). If one combination seems superior in terms of accuracy rate, shall we adopt it as the one to use or is there a combination of the approaches that results in some benefit? We show here how we have combined classifiers considering the misclassification cost to obtain a more informative classification for its application in the field.