A support vector machine-based analysis method with wavelet denoised near-infrared spectroscopy

This paper is concerned with the rapid and non-destructive quantitative analysis of cimetidine in single intact tablets by diffuse reflectance spectroscopy. Support vector machines (SVM) are introduced to model multivariate, non-linear systems of calibration samples by radical basis functions. Short-wave near-infrared spectra ranging 760–1100 nm are processed by SVM. Wavelet method has been employed to minimize the influence of noise. Measurement errors of independent testing set by SVM compared to partial least squares (PLS) give relatively reasonable results. Experiments show that SVM with wavelet denoising pretreatment is an effective method and requires less number of calibration samples.