The essential oils extracted from three kinds of herbs were separated by a 5% phenylmethyl silicone (DB-5MS) bonded phase fused-silica capillary column and identified by MS. Seventy-four of the compounds identified were selected as origin data, and their chemical structure and gas chromatographic retention times (RT) were performed to build a quantitative structure–retention relationship model by genetic algorithm and multiple linear regressions analysis. The predictive ability of the model was verified by internal validation (leave-one-out, fivefold, cross-validation and Y-scrambling). As for external validation, the model was also applied to predict the gas chromatographic RT of the 14 volatile compounds not used for model development from essential oil of Radix angelicae sinensis. The applicability domain was checked by the leverage approach to verify prediction reliability. The results obtained using several validations indicated that the best quantitative structure–retention relationship model was robust and satisfactory, could provide a feasible and effective tool for predicting the gas chromatographic RT of volatile compounds and could be also applied to help in identifying the compound with the same gas chromatographic RT.