基于最小二乘支持向量机的小样本建模方法研究

针对捷联惯组历次测试数据小样本建模问题,提出了通过二次修正插值方法解决测试数据的非等间隔性和样本容量小的问题。并通过相空间重构的思想将一维时间序列多维化。最后通过最小二乘支持向量机建立预测模型。实例分析表明,建立在二次修正插值基础之上的最小二乘支持向量机时间序列模型具有较高的预测精度,能够很好地满足对惯组测试数据分析的要求。

Time Series Prediction Based on LS-SVM

Aiming at the small sample model building problem of SIMU(Strapdown Inertial Measurement Unit) successive test data,a double-modified-interpolation method is provided to solve the unequal interval problem and small capacity problem of the sample.Phase-space re-building method is used to extend the time series from one-dimension to many-dimension.LS-SVM(least square support vector machines) method is presented to build the forecasting model.The illustration shows that LS-SVM forecasting model based on the double-modified-interpolation method has good performance and high precision.