Abstract: Traditional phenomenological cure kinetics models often exhibit limited compatibility with the complex reaction mechanisms involved in the curing process of thermosetting composite systems. In this work, a cure kinetics model for an interlaminar-toughened epoxy composite was developed using a Bagged decision trees machine learning algorithm. Modulated differential scanning calorimetry (mDSC) was employed to separate reversible and irreversible heat flow components, thereby eliminating the interference of thermoplastic particle melting endotherms on the characterization of the epoxy curing exotherm. A comprehensive cure dataset was constructed incorporating multiple heating rates and initial degrees of cure. The Bagged decision trees model was utilized to train multiple decision-tree sub-models via random sampling with replacement, which were subsequently integrated through averaging or majority voting to achieve robust cure kinetics modeling. To evaluate model accuracy, the prediction performance of the machine learning model was compared with that of conventional phenomenological models—including the autocatalytic model, the dual-reaction mechanism model, the diffusion-controlled dual reaction mechanism model and the dif-fusion-controlled autocatalytic model—as well as with experimental results. The results indicated that the machine learning model exhibited excellent agreement in overall fitting behavior and predictive accuracy, with a root-mean-square error (RMSE) as low as 0.000012032 and a coefficient of determination (R²) reaching 0.9998. This study demonstrates that the proposed machine learning approach eliminates the need for prior assumptions regarding reaction mechanisms, offering a generalized cure kinetics modeling framework with strong compatibility and transferability.