Eventos Anais de eventos
COBEM 2017
24th ABCM International Congress of Mechanical Engineering
Investigating the use of random forest, gradient boosting machine, support vector machine and their ensemble applied to fault detection
Submission Author:
Leandro dos Santos Coelho , PR , Brazil
Co-Authors:
Luis Felipe Nogoseke, Gabriel Herman Bernardim Andrade, Marco Boaretto, Leandro dos Santos Coelho
Presenter: Luis Felipe Nogoseke
doi://10.26678/ABCM.COBEM2017.COB17-1600
Abstract
The idea of ensemble methodology is to build a predictive model by integrating multiple models. It is well-known that ensemble methods (called ensemble classifiers or multiple classifier systems) can be used for improving prediction performance in classification tasks. In recent years, machine learning and pattern recognition approaches have been a promising tool in the field of fault diagnosis. The fault detection is an approach to catch any abnormal events of the system quickly or in advance and notify the problems to the system user. The main contribution of this paper is to verify the effectiveness of using ensemble learning combining extremely randomized trees, gradient boosting machine, k-nearest neighbors and artificial neural networks for fault detection case studies related to signal processing and mechanical systems. The ensemble approach with a stacking method reached a better performance when compared with the other machine learning techniques applied alone in both of the two fault study cases problems. Proving that the combination of different techniques with low accuracy scoring, can result in a final model that overachieves the overall performance of the techniques applied alone.
Keywords
machine learning, Random Forest, Gradient boosting machine, Support vector machine, Fault Detection
