Method to improve discrimination using movingwindow principal component analysis (MWPCA) for origin of products

Abstract

A new discrimination method called the moving-window principal component analysis (MW-PCA) has been developed and its performance has been evaluated using several spectroscopic datasets. The main concept of MW-PCA was to combine moving-window system and principal component analysis (PCA), and then using an effective algorithm of error rate to obtain a value of percent of a separation. This algorithm has been using a standard deviation of each groups to make a separation line. To evaluate its discrimination performances, four different spectroscopic datasets were employed: (1) conventional Raman spectra and wide area illumination Raman spectra of a origin of rice, (2) near-infrared(NIR) of a origin of cnidium officinale, (3) near-infrared(NIR) of a origin of carrot, (4) near-infrared(NIR) of a origin of sesame. For each case, results of separation were achieved. Since the method of MW-PCA is different from other algorism and great to separate groups. Combining moving-window and principal component analysis with using newly algorism of error rate provided better result of qualitative analysis.