Abstract: Events with highly similar waveforms and rupturing of the same fault patch are interpreted as repeating earthquakes, which can be applied in detecting deep fault deformation, characterizing fault behavior, and assess seismic hazards. In this study, we develop a clustering-based repeating earthquakes identification method by using the hierarchical clustering algorithm in machine learning. First, the parallel waveform cross-correlation method is adopted to calculate the cross-correlation coefficient (CC). Then, the S-P differential time is used to measure the inter distance of events. Finally, the hierarchical clustering is applied to obtain repeating clusters. We utilize this method to investigate the seismicity around the Ganzi-Yushu fault (GYF) and the eastern Kunlun fault (EKLF). We identify 6 repeating clusters along the GYF, with an average fault slip rate of 7.4 mm/a. Around the EKLF, we identify 3 repeating clusters, with an average fault slip rate of 6.9 mm/a. Along the EKLF, the fault slip rates gradually decreases from the west to the east along the strike, indicating a complex dynamic process. Our results agree with geology observation and GPS data. Based on real data testing, our results show that the method to identify repeating earthquakes is automatic, efficient and convenient and provides basic information for accurate identification of repeating earthquakes and places constraints on fault activity.