A novel optimization algorithm integrating immunity clone and differential evolution for parameter selection of SVM

作者：Li Jun^*; Ding Lixin; Zou Qilin

来源：Information Technology Journal, 2013, 12(23): 7367-7372.

DOI：10.3923/itj.2013.7367.7372

• Classification accuracy
• Clone selection
• Differential evolution algorithms
• Global searching ability
• Hybrid optimization algorithm
• Optimization algorithms
• Parameter optimization
• Pre-mature convergences

The classification accuracy of Support Vector Machine (SVM) depends on parameters strongly.In nature, parameter selection is a search optimization process. A Differential Evolution (DE) algorithm is a realcoding optimal algorithm based of swarm evolution. It has powerful global searching ability. But it gets into premature convergence easily. So a novel hybrid optimization algorithm based on Immunity Clone (IC) and differential evolution is proposed for parameter selection of SVM. In this algorithm, clonal selection and receptor edit mechanism are inserted into the differential evolution process. Thirteen experimental results on UCI datasets distinctly show that compared with default parameters SVM classifier, the differential evolution algorithm, the proposed algorithm has higher classification accuracy.

• 出版日期2013
• 单位武汉科技大学