The molecular electronegativity-distance vector (MEDV) descriptors are used to describe the structures of 69 compounds isolated from the essential oil of Paulownia tomentosa flowers. Through multiple linear regression (MLR), the correlation coefficient 0.964 is achieved. We select five and six vectors to build the models through the method of stepwise multiple regression (SMR), and the correlation coefficients of the two models are 0.962 and 0.964. Then all the models are evaluated by performing the cross-validation with the leave-one-out (LOO) procedure and the satisfactory results with correlation coefficients of 0.941, 0.954 and 0.954, respectively. So we can get the conclusion that the stability and predictability of the model is good.