Please use this identifier to cite or link to this item:
|標題:||Hierarchical Cluster-Based Multispecies Particle-Swarm Optimization for Fuzzy-System Optimization||作者:||Juang, C.F.
|關鍵字:||Fuzzy modeling;fuzzy prediction;particle-swarm optimization (PSO);swarm intelligence;multiple data sets;statistical comparisons;genetic algorithm;neural-network;convergence;design;identification;classifiers;prediction;strategy||Project:||Ieee Transactions on Fuzzy Systems||期刊/報告no：:||Ieee Transactions on Fuzzy Systems, Volume 18, Issue 1, Page(s) 14-26.||摘要:||
This paper proposes a hierarchical cluster-based multispecies particle-swarm optimization (HCMSPSO) algorithm for fuzzy-system optimization. The objective of this paper is to learn Takagi-Sugeno-Kang (TSK) type fuzzy rules with high accuracy. In the HCMSPSO-designed fuzzy system (FS), each rule defines its own fuzzy sets, which implies that the number of fuzzy sets for each input variable is equal to the number of fuzzy rules. A swarm in HCMSPSO is clustered into multiple species at an upper hierarchical level, and each species is further clustered into multiple subspecies at a lower hierarchical level. For an FS consisting of r rules, r species (swarms) are formed in the upper level, where one species optimizes a single fuzzy rule. Initially, there are no species in HCMSPSO. An online cluster-based algorithm is proposed to generate new species (fuzzy rules) automatically. In the lower layer, subspecies within the same species are formed adaptively in each iteration during the particle update. Several simulations are conducted to verify HCMSPSO performance. Comparisons with other neural learning, genetic, and PSO algorithms demonstrate the superiority of HCMSPSO performance.
|Appears in Collections:||電機工程學系所|
Show full item record
TAIR Related Article
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.