Recently, genetic engineering efforts have been made to develop recombinant Saccharomyces cerevisiae strains able to utilize xylose, an inexpensive and abundant carbon source. However, their construction and selection processes are limited by the speed and expenses of the existing testing methods, thus a rapid and equally precise method will significantly increase the number of tested strains. Here, near infrared (NIR) spectroscopy is proposed as a successful alternative method for screening recombinant xylose-fermenting S. cerevisiae. Supernatant samples of fermentation solutions from one diploid and three haploid recombinant strains were collected along the fermentation process. NIR spectra of the diluted supernatant provided effective differentiation of strains consistent with their phenotypic and genotypic features. This result could be used as a feedback for multicomponent analysis, in order to develop regression model for quantification of consumed glucose and xylose, produced ethanol, glycerol, and xylitol. Robust partial least-squares regression models with high prediction accuracy that are effective with any strain were achieved for all components when the modeling was performed with combined data of all strains, achieving 0.21-1.49 g/L of standard error of prediction with calibration, prediction, limit of detection and limit of quantification in the range of 1.0-4.5 and 3.0-13.4 g/L, respectively.