Continuous cooling transformation (CCT) diagram is an important basis to make an optimal heat treatment process of steels with a desired microstructure and properties. Therefore, it is of great practical importance to predict the CCT diagram rapidly and accurately, especially for costly and time-consuming material design by trial and error. In this study, machine learning approaches are provided to predict CCT diagrams of tool steels using relevant material descriptors including the chemical compositions, austenitizing temperature and cooling rate. Different machine learning techniques including the Multilayer Perceptron Regressor, k-Nearest Neighbours, Bagging and Random Forest are performed on experimental dataset for appropriate model selecting. Random forest is proved to be the best model to predict pearlite transition temperature and martensite transformation start temperature accurately. K-Nearest Neighbours and Bagging are suitable for predicting the start and end temperatures of bainite formation respectively. These optimal models are then used to predict the CCT diagrams of T8, 6CrW2Si, 4CrMoV, CrMn and Cr12W. Comparing with the calculation results by the commercial software JMatPro, these optimal models work much better for their distinguished prediction performance with high correlation coefficient and low error values, which is thus of great engineering significance.