Prediction of active substance content in pharmaceutical tablets using wavelet compressed NIR and Raman spectroscopic data

Wavelet transformation is performed on NTR transmittance and Raman spectroscopic data followed by prediction of active substance content of pharmaceutical tablet from the spectral data. Partial least squares regression (PLSR) is used to build the prediction models. Comparison is made between prediction models with and without wavelet compression. Results show that wavelet-transformed NIER spectral data extracted from certain wavelet scales could improve the prediction accuracy of the PLSR model. In addition, wavelet transformation of NIR spectral data reduces the number of variables used in calibration and prediction. However, there is no improvement in prediction accuracy for wavelet-transformed Raman spectral data.