Author:

Keywords:

Science & Technology, Physical Sciences, Chemistry, Analytical, Chemistry, Firefly algorithm, Variable selection, Chemometrics, Spectroscopy, NEAR-INFRARED SPECTROSCOPY, LEAST-SQUARES REGRESSION, MULTIVARIATE CALIBRATION, GENETIC ALGORITHMS, CHEMOMETRICS, OPTIMIZATION, MODELS, INFORMATION, QUALITY, PROTEIN, DESIGN, FOOD, Algorithms, Animals, Artificial Intelligence, Calibration, Fireflies, Least-Squares Analysis, Spectroscopy, Fourier Transform Infrared, 0301 Analytical Chemistry, 0399 Other Chemical Sciences, Analytical Chemistry, 3401 Analytical chemistry, 4004 Chemical engineering, 4018 Nanotechnology

Abstract:

A critical step in multivariate calibration is wavelength selection, which is used to build models with better prediction performance when applied to spectral data. Up to now, many feature selection techniques have been developed. Among all different types of feature selection techniques, those based on swarm intelligence optimization methodologies are more interesting since they are usually simulated based on animal and insect life behavior to, e.g., find the shortest path between a food source and their nests. This decision is made by a crowd, leading to a more robust model with less falling in local minima during the optimization cycle. This paper represents a novel feature selection approach to the selection of spectroscopic data, leading to more robust calibration models. The performance of the firefly algorithm, a swarm intelligence paradigm, was evaluated and compared with genetic algorithm and particle swarm optimization. All three techniques were coupled with partial least squares (PLS) and applied to three spectroscopic data sets. They demonstrate improved prediction results in comparison to when only a PLS model was built using all wavelengths. Results show that firefly algorithm as a novel swarm paradigm leads to a lower number of selected wavelengths while the prediction performance of built PLS stays the same.