Abstract: Based on 182 experimental data of austenite decomposition temperature,the formulas or models of austenite decomposition temperature are established by linear regression method,nonlinear regression method and artificial neural network respectively.The results show that linear regression formula cannot exactly express the relationship between factors and austenite decomposition temperature,with molybdenum exponential and ln(t_8/3) function were introduced,precision of the formula increased; the neural network model takes the interaction of factors into account,which predict more accurately than by linear and nonlinear regression formulas,and is more applicable for the research of predicting austenite decomposition temperature.