论文部分内容阅读
对飞机噪声距离特性(NPD)回归模型的选用进行了研究。通过NPD曲线特征分析给出了线性模型、多项式模型、神经网络和支持向量机(SVM)等四种备选回归模型,结合F16战斗机NPD数据实例对比研究了各种模型的回归与预测效果。结果表明,SVM模型误差最小,性能最稳定,其它模型回归内插精度较高,但外推预测误差较大,性能不够稳定。
The selection of aircraft noise distance characteristic (NPD) regression model was studied. Four candidate regression models of linear model, polynomial model, neural network and support vector machine (SVM) are given through NPD curve characteristic analysis. The regression and prediction results of various models are compared with the NPD data of F16 fighter. The results show that the SVM model has the smallest error, the most stable performance and the other models have higher regression interpolation precision, but the extrapolation prediction error is larger and the performance is not stable enough.