论文部分内容阅读
为克服光谱分析中异常训练样本的影响,提出了一种加权最小二乘支持向量机(WLS-SVM)的稳健化迭代算法.针对原始WLS-SVM在收敛性和稳健性方面的不足,提出了一种新的求取回归误差的方法,从而从根本上解决了WLS-SVM的收敛性问题;同时对原始算法求权值的步骤进行了修正,采用回归误差的中值作为计算加权值的比较基准,大幅度提高了WLS-SVM的稳健性.将算法应用于光谱定量分析中,实验结果证明了该方法是收敛的,并且崩溃点在35%左右,是一种有效的稳健建模方法.
In order to overcome the influence of abnormal training samples in spectral analysis, a robust iterative algorithm based on weighted least squares support vector machine (WLS-SVM) is proposed. In order to overcome the shortcomings of the original WLS-SVM in terms of convergence and robustness, A new method to find the regression error is to solve the convergence problem of WLS-SVM fundamentally. At the same time, the steps of calculating the weight value of the original algorithm are modified, and the median of the regression error is used as a comparison of the weighted values Which greatly improves the robustness of WLS-SVM.The experimental results show that the proposed method is convergent and has a collapse point of about 35%, which is an effective robust modeling method.