Parameter selection method for support vector machine based on adaptive fusion of multiple kernel functions and its application in fault diagnosis

• Published in: Neural computing & applications Vol. 32; no. 1; pp. 183 - 193
• Main Authors: Wang, Hailun, Xu, Daxing, Martinez, Alexander
• Format: Journal Article
• Language: English
• Published: London Springer London 01.01.2020; Springer Nature B.V
• Subjects: Adaptive filters; Artificial Intelligence; Brain- Inspired computing and Machine learning for Brain Health; Computational Biology/Bioinformatics; Computational Science and Engineering; Computer Science; Computer simulation; Data Mining and Knowledge Discovery; Estimating techniques; Fault diagnosis; Image Processing and Computer Vision; Kalman filters; Kernel functions; Mathematical models; Nonlinear systems; Parameter estimation; Probability and Statistics in Computer Science; Regression analysis; State estimation; State vectors; Support vector machines; Parameter selection; Support vector regression machine; Fifth-degree cubature Kalman filter; Kernel function characteristics
• ISSN: 0941-0643; 1433-3058
• DOI: 10.1007/s00521-018-3792-7

A new model parameter selection method for support vector machine based on adaptive fusion of multiple kernel functions is proposed in this paper. Characteristics of local kernels, global kernels, mixtures of kernels and multiple kernels were analyzed. Fusion coefficients of the multiple kernel function, kernel function parameters and regression parameters are combined to form the parameters of the state vector. Thus, the model selection problem is transformed into a nonlinear system state estimation problem. Then, we use a fifth-degree cubature Kalman filter to estimate the parameters. In this way, we realize adaptive selection of the multiple kernel function weighted coefficient, the kernel parameters and the regression parameters. A simulation experiment was performed to interpret the PE process for fault diagnosis.