Support Vector Machine Optimized by Henry Gas Solubility Optimization Algorithm and Archimedes Optimization Algorithm to Solve Data Classification Problems

Support vector machine (SVM) is the minimization of structural risk to construct a better hyperplane to maximize the distance between the hyperplane and the sample points on both sides of hyperplane. Two improved physics-wise swarm intelligence optimization algorithms (Henry gas solubility optimizat...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Engineering letters Ročník 31; číslo 2; s. 531
Hlavní autoři: Yu, Ji-Sheng, Zhang, Sheng-Kai, Wang, Jie-Sheng, Li, Song, Sun, Ji, Wang, Rui
Médium: Journal Article
Jazyk:angličtina
Vydáno: Hong Kong International Association of Engineers 23.05.2023
Témata:
Algorithms
Brownian motion
Classification
Datasets
Gas solubility
Hyperplanes
Kernel functions
Optimization
Support vector machines
Surface defects
Swarm intelligence
ISSN:1816-093X, 1816-0948
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Support vector machine (SVM) is the minimization of structural risk to construct a better hyperplane to maximize the distance between the hyperplane and the sample points on both sides of hyperplane. Two improved physics-wise swarm intelligence optimization algorithms (Henry gas solubility optimization algorithm and Archimedes optimization algorithm) were proposed based on Lévy flight operator, Brownian motion operator and Tangent flight motion operator to optimize the penalty factor and kernel function parameters of SVM so as to enhance its global and local search ability. Finally, the Iris datasets, Strip surface defect datasets, Wine datasets and Wisconsin datasets of breast cancer in UCI datasets were selected to carry out the simulation experiment. Simulation results show that optimizing SVM based on improved physical-wise swarm intelligence algorithms can effectively improve the classification accuracy.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1816-093X
1816-0948