Abstract: 3D-Quantitative Structure-Activity Relationship (3D-QSAR) genetic function approximation model was established to predict the sweetness of some chemical constituents in cigarette smoke by quantum chemistry calculation and statistics method.Among them, 30 compounds were used as training set and 10 compounds as validate set.The results showed that it was the higher prediction accuracy in this model, and the coefficient of correlation between experimental values and predict value was up to 0.9538 in the validation of 10 compounds.The model supplied theoretical basis on predicting sweetness of compounds in cigarette smoke.