Título: Análise Comparativa entre Métodos Multivariados no Problema de Quantificação do Princípio Ativo Hidroclorotiazida em Amostras de Medicamentos Anti-hipertensivos
Autores: Molz, Rolf Fredi; Konzen, Pedro H. A.; Furtado, João Carlos; Ferrão, Marco Flores; Carvalho, Claudia Willke; Crossetti, Geraldo Lopes
Resumo: This work presents a comparative analysis between the multi-variate regression method by Partial Least Squares (PLS) and the Artificial Neural Networks (ANN) in the active principle quantification problem to samples of anti-hyphertensive pharmacological compounds. In the first experiment, two techniques were used to understand the analite quantification: Diffuse Reflection in Infrared Fourier Transform Spectroscopy (DRIFTS) and PLS. The second experiment involves the use of multi level perceptron algorithm with backpropagation training in the PLS substitution. Both experiments used the same DRIFTS data set. The implementation and the calibration models were developed in MATLAB environment. The training and validation spectra sets were obtained by a diffuse reflectance accessory joined to a Nicolet Magna 550 spectrometer on the medium infrared region.
Código DOI: 10.21528/CBRN2003-030
Artigo em PDF: 6CBRN_030.PDF
Arquivo BibTex: 6CBRN_030.bib