Abstarct: In the first section of this project, quantitative structure - property relationship (QSPR) has been developed for modeling and prediction of retention indices of 60 volatile organic compounds (VOCs). The stepwise regression (SR) and genetic algorithm (GA) were used as variable selection methods. The selected variable by these methods was used as input of artificial neural network (ANN) to construct of model to predict of retention indices of these compounds. The obtained models were validated by different techniques such as using the external test set, leave one out (LOO) and y-randomization. The obtained results Shaw that both models of SR-ANN and GA-ANN have a good prediction ability. The determination coefficient of the test set for SR-ANN and GA-ANN were obtained 0.995, 0.992 respectively. In the second part of this project, a QSPR model has been developed to predict of retention indices of some sterol compounds, same the first part, SR and GA were used as variable selection methods. Also, ANN technique was applied for modeling and prediction of retention indices of these compounds. The obtained results of determination coefficient of the test set for SR-ANN and GA-ANN were obtained 0.944, 0.920 respectively.