To meet the growing demand for wireless data transmission, heterogeneous network technologies deploy numerous low-power nodes within macro base stations, forming a macro-micro cellular network to provide extended coverage. However, co-deployment of macro and micro base stations on the same frequency often leads to strong interference from macro base stations affecting edge users of micro base stations. Effective interference coordination methods are therefore necessary to enhance system throughput. This paper proposes a MIMO beamforming algorithm for macro-micro heterogeneous networks based on heterogeneous interference graph neural network (HIGNN) and clustering scheduling. Initially, the heterogeneous network is clustered based on interference intensity, followed by decision-making using the HIGNN-based beamforming algorithm to determine the optimal beamforming vectors for transmission in high-interference clusters. Simulation results show that the proposed algorithm reduces computational overhead effectively and improves system throughput.