Induction welding for thermoplastic composite materials uses an alternating current flowing through a coil to induce an electromagnetic field and generate eddy current inside laminate with various fiber orientations-the generated heat causes the laminate to heat up and melt the polymer. As a pressure is applied to the induction heating zones, cohesive bonding may occur during the melting of the polymer. The welding quality of the composite materials is highly influenced by the temperature varying inside the heating zones. Thus, it is beneficial for induction welding if temperature varying during heating can be acquired given a set of welding parameters, such as current, pressure, fiber orientations, etc. Conducting practical induction heating experiments for this purpose is laborious and time consuming given the large varying space of welding parameters. In this paper, we propose to address this problem by using machine learning techniques to model the relation between the welding parameters and the temperature varying inside the heating zones. We conduct two sets of induction heating experiments for laminate welding and the collected sample temperature varying data are used to train the neural networks with input of welding parameters and output of the predicted temperature varying. Testing of the models demonstrates that process modeling of induction welding with machine learning techniques is viable.