Feature selection algorithms using Chilean wine chromatograms as examples

This work presents the results of applying genetic algorithms, in selecting the more relevant features present in chromatograms of polyphenolic compounds, obtained from a high performance liquid chromatograph with aligned photodiodes detector (HPLCDAD), of samples of Chilean red wines Cabernet Sauvignon, Carmenere and Merlot. From the 6376 points of the original chromatogram, the genetic algorithm is able to select 37 of them, providing better results, from classification point of view, than the case where the complete information is used. The percent of correct classification reached with these 37 features turned out to be 94.19%.