Development of genetic algorithm based classification and cluster analysis methods for analytical data

Abstract

In this study genetic algorithm based classification and clustering methods were aimed to develop for the spectral data. The developed methods were completely achieved hybridization of nature inspired algorithm (genetic algorithms, GAs) to other classification or clustering methods. The first method was genetic algorithm based principal component analysis (GAPCAD), and the second was genetic algorithm based discriminant analysis (GADA). Both methods were performed to achieve the best discrimination between the olive oil and vegetable oil samples. The classifications of samples were examined directly from their spectral data obtained from using near infrared spectrometry, Fourier transform infrared (FTIR) spectrometry, and spectrofluorometry. The GA was used to optimize the performance of classification or clustering techniques. on training set in order to maximize the correct classification of acceptable and unacceptable samples or samples of dissimilar properties and to reduce the spectral data by wavelength selection. After GA optimization the classification results of training set were controlled by validation set. Lastly, the success of both algorithms was compared to the results of PCA and SIMCA.