Dynamic data modeling based on multi-kernel support vector machine with immune sliding optimization

A novel dynamic modeling method based on multi-kernel support vector machine with immune sliding optimization (ISA-MKSVM) is proposed in this paper. Multi-kernel SVM is an effective tool to solve the diversity of data features, and the immune sliding algorithm optimizes the parameters of multi-kernel SVM. The proposed algorithm can obtain a good dynamic performance by re-defining the affinity of the antibody on a dynamic system. The proposed approach is applied for modeling the solvent concentration in forming of carbon fiber, and the experimental results show that ISA-MKSVM is accurate and has better generalization ability than single-kernel SVM, and the process of parameter optimization is faster than traditional algorithm.