基于改进网格搜索法的SVM参数优化
|
摘要
支持向量机的学习和泛化能力很大程度上取决于其相关参数的选取.针对传统网格算法的不足,引入全局粒子群算法,利用其能够快速到达最优解附近的优势:先使用粒子群算法进行粗搜;再使用网格搜索法进行小步长的精细搜索得到最优解.实验结果表明:基于改进的网格搜索法SVM对比传统网格搜索法SVM,在预测精度和运算时间上都具有优势.
The learning and generalization ability of SVM depends on the selection of related parameters.Considering the deficiency of the traditional grid search method, this paper introduces particle swarm optimization(PSO) to optimize the algorithm which can rapidly reach the optimal solution. This paper first uses PSO to search the optimum solution; then uses grid search method for fine search with small length to get the final solution. The results show that the new algorithm has better prediction accuracy and shorter computation time compared with the traditional grid method.
The learning and generalization ability of SVM depends on the selection of related parameters.Considering the deficiency of the traditional grid search method, this paper introduces particle swarm optimization(PSO) to optimize the algorithm which can rapidly reach the optimal solution. This paper first uses PSO to search the optimum solution; then uses grid search method for fine search with small length to get the final solution. The results show that the new algorithm has better prediction accuracy and shorter computation time compared with the traditional grid method.
引文
[1]Vapnik V,Levin E,Cun Y L.Measuring the VC-dimension of learning machines[J].Neural Computation,1994,6(5):851-876.
[2]邓乃扬,田英杰.支持向量机---理论、算法与拓展[M].北京:科学出版社,2009.
[3]李喆,柏丛,孙健,等.基于PSO-SVM的出行方式识别研究[J/OL].计算机应用研究,2016(12):1-5.
[4]曲健,陈红岩,刘文贞,等.基于改进网格搜索法的支持向量机在气体定量分析中的应用[J].传感技术学报,2015,28(5):774-778.
[5]郑志成,徐卫亚,徐飞,等.基于混合核函数PSO-LSSVM的边坡变形预测[J].岩土力学,2012,33(5):1421-1426.
[6]刘东平,单甘霖,张岐龙,等.基于改进遗传算法的支持向量机参数优化[J].微计算机应用,2010,31(5):11-15.
[7]秦玲.蚁群算法的改进与应用[D].扬州:扬州大学,2004.
[8]俞俊平,陈志坚,武立军,等.基于蚁群算法优化支持向量机的边坡位移预测[J].长江科学院院报,2015,32(4):22-27.
[9]Vapnik V N.The nature of statistical learning theory[M].New York:Springer-Verlag,1995.
[10]徐飞,徐卫亚,王珂.基于蚁群优化支持向量机模型的边坡稳定性[J].工程地质学报,2009,17(2):253-257.
[11]赵洪波,冯夏庭.非线性位移时间序列预测的进化-支持向量机方法及应用[J].岩土工程学报,2003,25(4):468-471.
[12]王健峰,张磊,陈国兴,等.基于改进的网格搜索法的SVM参数优化[J].应用科技,2012,39(3):28-31.
[13]J Kennedy,R Eberhart.Particle swarm optimization[C]//Neural Networks,1995.Proceedings,IEEE International Conference on.IEEE,1995:1942-1948.
[14]毛伊敏,周昭飞,彭喆,等.基于不确定多分类支持向量机在滑坡危险性预测的应用[J].江西理工大学学报,2016,37(3):102-108.
[15]范永东.模型选择中的交叉验证方法综述[D].太原:山西大学,2013.
[2]邓乃扬,田英杰.支持向量机---理论、算法与拓展[M].北京:科学出版社,2009.
[3]李喆,柏丛,孙健,等.基于PSO-SVM的出行方式识别研究[J/OL].计算机应用研究,2016(12):1-5.
[4]曲健,陈红岩,刘文贞,等.基于改进网格搜索法的支持向量机在气体定量分析中的应用[J].传感技术学报,2015,28(5):774-778.
[5]郑志成,徐卫亚,徐飞,等.基于混合核函数PSO-LSSVM的边坡变形预测[J].岩土力学,2012,33(5):1421-1426.
[6]刘东平,单甘霖,张岐龙,等.基于改进遗传算法的支持向量机参数优化[J].微计算机应用,2010,31(5):11-15.
[7]秦玲.蚁群算法的改进与应用[D].扬州:扬州大学,2004.
[8]俞俊平,陈志坚,武立军,等.基于蚁群算法优化支持向量机的边坡位移预测[J].长江科学院院报,2015,32(4):22-27.
[9]Vapnik V N.The nature of statistical learning theory[M].New York:Springer-Verlag,1995.
[10]徐飞,徐卫亚,王珂.基于蚁群优化支持向量机模型的边坡稳定性[J].工程地质学报,2009,17(2):253-257.
[11]赵洪波,冯夏庭.非线性位移时间序列预测的进化-支持向量机方法及应用[J].岩土工程学报,2003,25(4):468-471.
[12]王健峰,张磊,陈国兴,等.基于改进的网格搜索法的SVM参数优化[J].应用科技,2012,39(3):28-31.
[13]J Kennedy,R Eberhart.Particle swarm optimization[C]//Neural Networks,1995.Proceedings,IEEE International Conference on.IEEE,1995:1942-1948.
[14]毛伊敏,周昭飞,彭喆,等.基于不确定多分类支持向量机在滑坡危险性预测的应用[J].江西理工大学学报,2016,37(3):102-108.
[15]范永东.模型选择中的交叉验证方法综述[D].太原:山西大学,2013.