【摘要】 支持向量回归是解决非线性时间序列预测问题的有效方法之一.为得到机场噪声时间序列预测的优化模型,将ν支持向量回归机解路径算法(ν-Svr Path)用到机场噪声时间序列的建模中,由此得到的模型在保证预测准确性的基础上,大大缩短了训练时间.在某机场噪声实测数据上的实验表明:由ν-Svr Path算法构造的预测模型在训练时间和预测的准确性上,其结果均优于盲目选取ν的模型;同时,由ν-Svr Path算法构造的模型的预测准确率高于ARMA和ε-SVR构造的模型,对机场噪声时间序列预测具有一定的实用价值.更多还原

【Abstract】 Support Vector Regression is one of effective methods for nonlinear time series prediction. This paper proposes using the existing solution ν-path algorithm( ν-Svr Path) to obtain an optimal time series model,which ensures the prediction accuracy and shortens the training time. The experiment on the measured data of airport noise shows that: The model based on ν-Svr Path was better than the model based on choosing ν arbitrarily in the prediction accuracy,while the training time was less. At the same time,the prediction accuracy of the model based on ν-Svr Path was better than the ARM A( autoregressive moving average) model and the ε-SVR model.The ν-path algorithm( ν-Svr Path) has practical value for the time series prediction modeling of the airport noise.更多还原