In order to reveal the relationship between flotation behaviors of collectors and their structures, quantitative structure–activity relationship (QSAR) study about separation efficiency of quartz from hematite using amine collectors was performed. The genetic function approximation (GFA) algorithm was applied to generate the correlation models and model with acceptable R2 and Rcv2 (cross validated R-squared) correlation coefficients (R2=0.9666, Rcv2=0.9201) was developed. The model revealed that the Lowest Unoccupied Molecular Orbital (LUMO) energy of the molecule, the charge of nitrogen and the electronegativity of polar group were the major factors that affected the separation efficiency of collectors. The higher nitrogen charge, the larger electronegativity of polar group and the more positive of LUMO energy of amine collectors were, the higher separation efficiency would be.