NIR spectrum of sample depends on the chemical composition and on the physical characteristics of the sample which are usually observed as the background and noise of the spectrum (Chen et al., 2007). Therefore, preprocessing options are used to eliminate offset or different linear baselines. Processing of the recorded average spectra data of Safflower seeds were carried out using the following preprocessing options: (1) no spectral data processing (2) constant offset elimination (3) straight line subtraction (4) vector normalization (5) min-max normalization (6) multiple scattering correction (7) first derivative (8) second derivative (9) first derivative + straight line subtraction (10) first derivative + vector normalization and (11) first derivative + multiple scattering. Preprocessing options were used to optimize the validation parameters in terms of rank, RMSECV and residual prediction deviation (RPD) values (Table 1).