Abstarct: Organo-sulfur compounds and poly cyclic aromatic hydrocarbons (PAHs) are known to be two groups of persistent pollutants. In addition, there is evidence of the carcinogenic potential of these compounds. There are different analytical techniques such as gas chromatography for separation and identification of these compounds by comparison of retention indices with standards. So, the retention index of a solute is an important parameter can be used for identification of organo-sulfur and PAHs compounds, but on the other hand, these methods are time consuming and very expensive. Therefore, to overcome on these problems, chemometrics methods can be used for calculation of retention indices of these compounds. In the first section of this thesis, a new method of Random Forests (RF) is used for predicting the retention indices of a group of sulfur compounds. Also the performance of the random forests model was compared with the artificial neural network (ANN) and multiple linear regressions (MLR). The determination coefficients obtained for test set by RF, SR-ANN, CDFS-ANN, RF-ANN and MLR are 0.997, 0.967, 0.994, 0.986 and 0.996 respectively. The results obtained showed that RF model has relative superiority than ANN and MLR to predict the retention indices of these compounds. In the second section, as same as first part, these methods were used to predict the retention indices of some PAH compounds. The determination coefficients obtained for test set by RF, SR-ANN, CDFS-ANN, RF-ANN and MLR are 0.997, 0.986, 0.989, 0.954 and 0.983 respectively. Obtained result showed that RF is better than the other techniques for prediction of RIs of PAHs.