Background: Partial least squares-1 (PLS) is a common simple, rapid and selective multivariate method for simultaneous determination of components having overlapping spectra in mixtures. Objectives: In the present study, a simple and sensitive chemometric procedure was suggested for the selective determination of alogliptin and pioglitazone without previous separation in pharmaceutical preparation using this powerful multivariate technique with and without variable selection (genetic algorithm). PLS method was run on the calibration data of absorption spectra. Pre-processing of the data was done and the regions below 215 nm and above 290 nm were rejected due to non-linearity, this results in 75 variable. Results: Genetic algorithm reduced absorbance matrix to about 61-56% of the original matrix (46 and 42 variables for alogliptin and pioglitazone, respectively). The selected variables were used for running the partial least squares model. Conclusion: A comparison between partial least squares and genetic algorithm- partial least squares models was done and the predictive ability of both models were evaluated.