Spectroscopic discrimination of Persian Gulf alga species using chemometrics approaches: A Comparative study using data fusion

Abstract

Abstract In the present study, multivariate chemometric techniques have been employed for discrimination of Persian Gulf seaweeds according to their species and area using FTIR and UV-Vis spectra. Based on the findings, this work could be differentiated diversity of algae species without time consuming sample preparation steps, measuring cell content and large consumption of material for analysis. The principal component analysis (PCA), PCA- discriminant analysis (PCA-LDA), and PLS-DA (partial least squares-discriminant analysis) were employed for clustering of samples. Based on our knowledge, the current work is the first study that examined the concatenation of SOMs clusters of different techniques for the classification of algae samples. Comparison of results showed that the classification parameters obtained by SOMs- fusion approach is better than other models and 100% prediction accuracy was achieved. The data sets were pretreated with autoscaling and Savitzky-Golay smoothing for elimination of the noise and background interferences and improving the accuracy of classification. Also, the Kohonen self-organization map (SOMs) was considered for reducing the dimension of spectroscopic data. The fusion matrix was created by a combination of best SOMs clusters of FTIR and UV-Vis spectra. The models are capable to distinguish the similarity between the face mask contain spirulina with spirulina alone.