In recent years, a great attention has been paid up on development of high manganese austenitic steels exhibiting high tensile strength and exceptional total elongation. Due to the low stacking fault energy (SFE), cross slip becomes more difficult in these steels and mechanical twinning is then the favored deformation mode beside of dislocation gliding. These alloys are therefore named as twinning-induced plasticity (TWIP) steels. On the other hand, twin boundaries act as strong obstacles for the subsequent movement of dislocations and affect mechanical properties, especially strain hardening rate. In this work, artificial neural network (ANN) models were developed in order to predict the process parameters affecting the tensile properties of high manganese austenitic TWIP steels. In these models, chemical composition (%Mn, %Al, %Si, and %C), cold rolling reduction, annealing/solution-treatment temperature and time, and strain rate were chosen as inputs. In all extracted data, other conditions such as achieving to TWIP steel (combination of large cold rolling reduction and subsequently annealing treatment in the partial recrystallization region), tensile test temperature and etc. were kept similar. The yield strength, engineering tensile strength and engineering total elongation were considered as outputs. The required databases for training and testing of these models were taken from some experiments as well as literature. All data were divided in two groups, 80% for training and 20% for testing. Both random and non-random (the last data) were used for the testing set. In order to validate the ANN models, several tensile tests were conducted under similar condition of cold rolling-annealing. To prepare the TWIP plates, several plates were cast, homogenized, cold rolled to 85% thickness reduction and subsequently annealed in the temperature range of 500–900°C for 30 min. Tensile tests were carried out with a strain rate of 10 -3 s -1 at room temperature. Specimens were characterized by X-ray diffraction, optical and scanning electron microscopy. The results showed better mechanical properties when annealed at temperature of 750 °C. Under this condition, a mixture of recrystallized and unrecrystallized regions with high density of mechanical twins was characterized. Based on the modeling results, a better correlation was found for the models with one single output instead of multiple outputs. A reasonable agreement was found between the results of tensile tests with the modeling predictions showing the robustness of the present ANN modes. Key Words High manganese steel, TWIP steel, Twinning, Stacking fault energy (SFE), Mechanical properties, Artificial neural network (ANN).