Independent Component Analysis and Support Vector Machine combined for Brands Identification of Milk Powder Based on Visible and Short-Wave Near-Infrared Spectroscopy

The aim of this paper is to investigate the potential of Visible and short-wave near-infrared spectroscopy (Vis/SWNIR) technique used for brand discrimination of milk powder. Fifty samples for each brand were studied. Based on the independent components (ICs) as the input variable obtained from fast fixed-point independent component analysis (FastICA), Least-squares Support Vector Machine (LS-SVM) was applied to building the prediction model. The discrimination rate of LS-SVM model which was established based on FastICA was reached at 100 %. LS-SVM model and partial least squares model, which were both established based on the whole measurement region, were also established. The identification results of these two models are worse than LS-SVM which was established based on ICs. It is concluded that Vis/ SWNIR technique is available for the brand identification of milk powder fast and non-destructively.