Aplicação de técnicas multivariadas e inteligência artificial na análise de espectros de infravermelho para determinação de matéria orgânica em amostras de solos

• Main Authors: Diego M. Souza, Beata E. Madari, Freddy F. Guimarães
• Format: Article
• Language: English
• Published: Sociedade Brasileira de Química 2012-01-01
• Series: Química Nova
• Subjects: artificial neural network; LS-SVM; soil organic matter
• Online Access: http://www.scielo.br/scielo.php?script=sci_arttext&pid=S0100-40422012000900007&lng=en&tlng=en

In this paper studies based on Multilayer Perception Artificial Neural Network and Least Square Support Vector Machine (LS-SVM) techniques are applied to determine of the concentration of Soil Organic Matter (SOM). Performances of the techniques are compared. SOM concentrations and spectral data from Mid-Infrared are used as input parameters for both techniques. Multivariate regressions were performed for a set of 1117 spectra of soil samples, with concentrations ranging from 2 to 400 g kg-1. The LS-SVM resulted in a Root Mean Square Error of Prediction of 3.26 g kg-1 that is comparable to the deviation of the Walkley-Black method (2.80 g kg-1).