For a long time, the heating pattern of the workpiece within a multimode microwave oven was considered to be highly sophisticated. As a consequence, the uneven microwave heating problem can only be partly alleviated by a random movement between the electromagnetic field and the workpiece. In this paper, we reported that the heating pattern has a specific correspondence with microwave system settings. The influence factor of the heating pattern and the corresponding mechanism were systematically studied by both theoretical analysis and experimental investigations. On this basis, a data-driven process model was established to learn the material’s dynamic temperature behaviors under different microwave system settings, and a new concept to improve the microwave heating uniformity by temperature monitoring and active compensation was proposed. The effectiveness of the method was demonstrated by a polymer composite microwave processing case study.