Feature extraction method based on LS-SVM and its application to intelligent quality control
一种基于LS-SVM的特征提取新方法及其在智能质量控制中的应用

A new feature extraction method based on Least Squares Support Vector Machine (LS-SVM) was proposed and applied to intelligent quality control successfully. Firstly, the formulation of linear feature extraction was made in the same fashion as that in the LS-SVM linear regression algorithm. Secondly, the data was mapped from the original input space to a high dimensional one by following the usual SVM methodology so as that nonlinear feature extraction can be made from linear version of the formulation through applying the kernel trick. Thirdly, 50 dimensional simulated data sets, including six patterns, extracted by universal control chart, were used to test. As a result, characteristics of the original data sets declined to be 3 dimensional and 80% classification-messages remained. Finally, the BP-based abnormal pattern recognizer was applied to the characteristics extracted samples, and better results were obtained than that of directly recognizing original samples with RSFM methods. The simulation results indicate that this feature extraction method is not only feasible but also effective.