基于和声搜索算法的支持向量机参数优化

支持向量机是建立在统计学理论基础上,以结构风险最小为原则的一种机器学习算法,能够很好地解决小样本、高维数、非线性等问题,被广泛地应用于模式识别、函数估计及回归预测等领域.但是支持向量机性能的高低往往取决于其相关参数的正确选择.为提高优化参数的精度及效率,利用和声搜索算法的全局寻优能力,对支持向量机的惩罚参数及核参数进行优化选择.通过4个标准UCI数据集的仿真实验,结果表明本算法不仅减少了搜索时间,而且所获得的参数能大幅提高支持向量机的性能和预测精度,提高了泛化能力.

Parameter Optimization of Support Vector Machine Based on Harmony Search Algorithm

The support vector machine method established on the basis of statistical theory is a machine learning algorithm of structural risk minimization principle, can solve the small sample, nonlinear problem, high dimentions etc. The performance of support vector machine depends on correct choice of the SVM parameters. The traditional parameter selection method is easy to fall into local optimum, in order to improve the precision and efficiency of the optimization parameters, the optimization algorithm of support vector machine parameter harmony search algorithm is adopted. Through the simulation of Matlab, the experimental results show that this algorithm not only decreases the search time, and by optimizing the parameters but also improves support vector machine prediction accuracy and generalization ability.