Thermophysical properties of oxygen are of great importance in practical applications. However, the values of the properties differs from each other under different circumstances, which may have bad influence in practical productions and applications. In our study, we mainly used computational models like Artificial Neural Networks (ANNs) to predict the thermophysical properties of the chemical substances. We succeeded in establishing 9 models to predict the thermophysical properties of oxygen, namely density, energy, enthalpy, entropy, isochoric heat capacity, isobaric heat capacity, viscosity and dielectric constant, by analyzing 51 data groups using linear prediction and Multilayer Feedfoward Neural Network (MLFN) methods. Within permissible error range (30% tolerance), all the tested samples were corresponded with the actual value. Our models were proved to be robust and accurate which indicated that ANN models can be applied in predicting the thermophysical properties of oxygen.