Support Vector Machines and Particle Swarm Optimization: Applications to Reliability Prediction - Couverture souple

Lins, Isis Didier; Das, Márcio; López, Enrique

 
9783838319407: Support Vector Machines and Particle Swarm Optimization: Applications to Reliability Prediction
Couverture souple
ISBN 10 : 3838319400 ISBN 13 : 9783838319407
Editeur : LAP Lambert Academic Publishing, 2010

Synopsis

Reliability is a critical indicator of organizations' performance in face of market competition, since it contributes to production regularity. Its prediction is of great interest as it may anticipate trends of system failures and thus enable maintenance actions. The consideration of all aspects that influence system reliability may render its modeling very complex and learning methods such as Support Vector Machines (SVMs) emerge as alternative prediction tools: previous knowledge about the function or process that maps input variables into output is not required. However, SVM performance is affected by parameters from the related learning problem. Suitable values for them are chosen by means of Particle Swarm Optimization (PSO), a probabilistic approach based on the behavior of organisms that move in groups. Thus, a PSO+SVM methodology is proposed to handle reliability prediction problems. It is used to solve application examples based on time series data and also involving data collected from oil production wells. The results indicate that PSO+SVM is able to provide competitive or even more accurate reliability predictions when compared, for example, to Neural Networks (NNs).

Les informations fournies dans la section « Synopsis » peuvent faire référence à une autre édition de ce titre.

Présentation de l'éditeur

Reliability is a critical indicator of organizations' performance in face of market competition, since it contributes to production regularity. Its prediction is of great interest as it may anticipate trends of system failures and thus enable maintenance actions. The consideration of all aspects that influence system reliability may render its modeling very complex and learning methods such as Support Vector Machines (SVMs) emerge as alternative prediction tools: previous knowledge about the function or process that maps input variables into output is not required. However, SVM performance is affected by parameters from the related learning problem. Suitable values for them are chosen by means of Particle Swarm Optimization (PSO), a probabilistic approach based on the behavior of organisms that move in groups. Thus, a PSO+SVM methodology is proposed to handle reliability prediction problems. It is used to solve application examples based on time series data and also involving data collected from oil production wells. The results indicate that PSO+SVM is able to provide competitive or even more accurate reliability predictions when compared, for example, to Neural Networks (NNs).

Les informations fournies dans la section « A propos du livre » peuvent faire référence à une autre édition de ce titre.

Éditeur
LAP Lambert Academic Publishing
Date d'édition
2010
Langue
anglais
ISBN 10 
3838319400
ISBN 13 
9783838319407
Reliure
Broché
Nombre de pages
92
Coordonnées du fabricant
non disponible
Personne responsable
KreGa Onlinehandel GmbH
4950000000000
info@krega-onlinehandel.de

Birkenweg 31
Brakel
33034
Allemagne