Multiple birth support vector machine based on dynamic quantum particle swarm optimization algorithm

At present, the parameters of the multiple birth support vector machine (MBSVM) are mainly determined by experience or artificially specified by the grid method. Both of these methods rely too much on the experience value, which is easy to cause the selection of the parameter value to be insufficien...

Ausführliche Beschreibung

Gespeichert in:

Bibliographische Detailangaben
Veröffentlicht in:	Neurocomputing (Amsterdam) Jg. 480; S. 146 - 156
Hauptverfasser:	Ding, Shifei, Zhang, Zichen, Sun, Yuting, Shi, Songhui
Format:	Journal Article
Sprache:	Englisch
Veröffentlicht:	Elsevier B.V 01.04.2022
Schlagworte:	Contraction expansion (CE) coefficient Dynamic quantum particle swarm optimization algorithm (DQPSO) Multiple birth support vector machine (MBSVM) Parameter optimization Multiple birth support vector machine (MBSVM) Contraction expansion (CE) coefficient Dynamic quantum particle swarm optimization algorithm (DQPSO) Parameter optimization
ISSN:	0925-2312, 1872-8286
Online-Zugang:	Volltext
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

Beschreibung
Zusammenfassung:	At present, the parameters of the multiple birth support vector machine (MBSVM) are mainly determined by experience or artificially specified by the grid method. Both of these methods rely too much on the experience value, which is easy to cause the selection of the parameter value to be insufficient. The selection of parameters directly affects the performance of MBSVM, so how to find a set of optimal parameters plays a vital role in improving the performance of MBSVM. To solve this problem, a dynamic quantum particle swarm optimization algorithm (DQPSO) was proposed. Focusing on the contraction – expansion (CE) coefficient control mode, authors propose the concept of the ability factor of particle search and realize the dynamic regulation of CE coefficient by taking it as feedback. Then DQPSO algorithm is used to optimize the parameters of MBSVM. The experimental results show that the proposed DQPSO algorithm has better optimization performance and faster convergence speed than that of the classical QPSO. Meanwhile, experiments on the UCI datasets show that the proposed DQPSO-MBSVM algorithm is effective to improve the classification performance of MBSVM.
ISSN:	0925-2312 1872-8286
DOI:	10.1016/j.neucom.2022.01.012