Abstarct: The main aim of this work is development of a quantitative structure–property relation (QSPR) method using an artificial neural network (ANN) for predicting the isothermal compressibility ( ) of different organic compounds. The data set consisted of 56 molecules in various temperatures and pressures, which form 4297 data point. The best set of calculated descxriptors was selected by two procedures: stepwise regression method and combined data splitting feature selection (CDFS) strategy. Modeling between selected descxriptors and was performed using ANN and after optimization and training of the network, it was used for the prediction of isothermal compressibility of seven external set compounds, which did not have contribute in model development steps. A comparison was made between two models. It was found that the CDFS model have lower number of descxriptors and higher prediction ability. The mean square error of the test set obtained by MLR-ANN and CDFS-ANN methods were 0.4456 and 0.3012, respectively. In the next step, artificial neural network and concept of group contribution method was simultaneously used to correlate and predict the of organic compounds. To do so, a correlation of functional group as well as temperature and pressure was used as input variables. After optimization the network, the model was used for prediction of of the same external set. The mean square error of the test set obtained by GCM-ANN method was 0.2303. Comparison of GCM-ANN mean square error of prediction values with those obtained using MLR-ANN and CDFS-ANN show the superiority of GCM-ANN over that MLR and CDFS models.