Wavelengths selection based on genetic algorithm (GA) and successive projections algorithms (SPA) combine with PLS regression for determination the soluble solids content in Nam-DokMai mangoes based on near infrared spectroscopy

The objective of this work was to search for an optimal wavelength selection for near infrared (NIR) spectroscopy for quality measurement of Nam-Dokmai mangoes. In this study, NIR spectroscopy has been applied to grading management systems for commercial mangoes export. Near infrared spectra were collected using a near infrared instrument incorporating a wavelength region of 860-1760 nm. Genetic algorithm (GA) and successive projections algorithms (SPA) was employed for selecting the spectra wavelengths. The selected wavelengths were also used to generate the prediction models via partial least square (PLS) regression. The optimal pretreatment was obtained from the second derivative. The model of full wavelengths rendered effective the best performance with r2 of 0.66-0.74, RMSEP of 0.72-0.80 °Brix and RPD equal to 1.8-2.0. The SPA-PLS resulted in values of r2, RMSEP and RPD were 0.43-0.70, 0.77-1.01°Brix and 1.4-1.9, respectively. Meanwhile, the result of GA-PLS performed efficiency with r2, RMSEP and RPD were 0.52-0.72, 0.74-0.96°Brix and 1.5-1.9, respectively. The outcome, the GA-PLS model (50 variables) is suitable for use in the measuring soluble solids content (SSC) in mangoes. This model could be used as screening purpose. It also was not different significantly when compared to the best model. Hence, authors suggested that the prediction model by GA-PLS with 50 variables can be effectively used for evaluating SSC in mangoes.