Using a genetic algorithm as an optimal band selector in the mid-near infrared: evaluation of the biodegradation of maize roots.

摘要: Mid- and near- infrared spectroscopies can provide useful information of the biomass composition. These methods have been extensively used in several applications such as biorefineries, biotechnologies, environment, especially for predicting plant composition or classify samples. However, numbers molecular descriptors mid- spectra are important sometimes redundant. Generally, only few spectral bands (wavenumbers) relevant applying regression classification models. The selection optimal subsets has addressed through including genetic algorithms. algorithms require an adapted fitness function order to identify most well suited bands. This study intends analyze influence functions both from a quantitative point view (which subset better describes biodegradation process) qualitatively highlights known chemical functional groups). Results obtained on recorded maize roots samples at periods process show that Davies-Bouldin function, which is measure separability between clusters within cluster scatter, gives best results.