Optimization of Principal Component Analysis in Feature Extraction

A novel method for optimising the principal component analysis in feature extraction is proposed, which makes use of parallel coordinate plot for graphical presentation of multivariate information. The objectivity and automatization of above manual observation and filtering process is realized by algorithm. In supervised multivariate information classification, before feature extraction on principal component analysis, filtering the variable that has bigger variance and has little effect on classification by observing the parallel coordinate plot of the multivariate data, the eigenvector from principal component analysis will be more in favor of classification. We achieved better performance when using this method to test the vegetable oil data. We believe that this method can be used in many other feature extraction methods, and will obtain better performance than them.