Grid-Search和PSO优化的SVM在Shibor回归预测中的应用研究
    点此下载全文
引用本文:张剑,王波.Grid-Search和PSO优化的SVM在Shibor回归预测中的应用研究[J].经济数学,2017,(2):84-88
摘要点击次数: 561
全文下载次数: 0
作者单位E-mail
张剑,王波 (上海理工大学 管理学院上海200093) Grid-Search和PSO优化的SVM在Shibor回归预测中的应用研究 
中文摘要:作为一种动态和非稳定时间序列,Shibor发展变化是随机波动的,难以准确预测Shibor的波动性.支持向量机(SVM)在回归预测非线性时间序列方面有很好地预测效果,SVM的预测精度和泛化能力的核心是参数的优化选择,分别用网格搜索法(Grid-Search)和粒子群(PSO)算法来优化SVM的参数c和g.从而将参数优化后的SVM非线性回归预测法与基于传统ARIMA时间序列预测结果进行对比分析.实验表明,优化后的SVM回归预测方法比ARIMA时间序列方法更精确,在实际中具有很大的应用价值.
中文关键词:机器学习  非线性回归预测  支持向量机  网格搜索法  粒子群算法  Shibor
 
Application of Grid-Search and PSO-optimized SVM in Shibor Regression Prediction
Abstract:As a dynamic and unsteady time series,the development of Shibor is a random fluctuation,and it is difficult to accurately predict the volatility of Shibor.Support vector machine (SVM) has a good predictive effect in the regression prediction of nonlinear time series.SVM’s prediction accuracy and generalization ability are due to the optimization of parameters.Grid-search and Particle Group (PSO) algorithm were used to optimize SVM parameters cand g.The SVM nonlinear regression prediction method with parametric optimization was compared with the traditional ARIMA time series prediction results.The experiments show that SVM regression prediction method is more accurate than ARIMA time series method,and it has great application value in practice.
keywords:machine learning  nonlinear regression prediction  support vector machines  grid-search algorithm  particle swarm optimization  Shibor
查看全文   查看/发表评论   下载pdf阅读器