基于遗传算法与经验误差最小化的SVM模型选择方法

SVM Model Selection Based on Genetic Algorithms and Empirical Error Minimization

支持向量机(SVM)的推广能力依赖于核函数形式及核参数和惩罚因子的选取,即模型选择.在分析参数对分类器识别精度的影响基础上,提出了基于遗传算法和经验误差最小化的支持向量机参数选择方法.在13个UC I数据集上的实验表明了本文算法的正确性与有效性,且具有良好的推广性能.

The spreading capacity of support vector machine （SVM） depends largely on the selection of kernel function and its parameters, and penalty factor, that is model selection. Having analyzed the parameter＇s influence on the classifier＇s recognition accuracy, we propose a new method for SVM model selection using genetic algorithm and empirical error minimization. The experiments on 13 different UCI benchmarks show its correctness, effectiveness and good spreading performance.