Abstract: Molecular electronegativity-distance vector (MEDV) was used to describe the chemical structure of 53 components of volatile components from Rosa banksiae Ait. Various multiple linear regression (MLR) models were created with variable screening by the stepwise multiple regression technique and statistics. The quantitative structure-retention relationship (QSRR) models of 10 variables, 6 variables and linear retention indices (ILR) of 10, 7, 6 variables were built by MLR with the correlation coefficients (R) of molecular modeling being 0.949, 0.906, 0.906, 0.949 and 0.943, respectively. Cross-validation of the models, which contained selected vectors, were performed by leave-one-out procedure (LOO) to obtain the satisfied results with correlation coefficients(RCV) of 0.867, 0.904, 0.905, 0.901 and 0.904, respectively. The results show that the constructed models possess good estimation stability and favorable predictive ability.