In this study, we cluster a large number of earthquake source time functions (STFs) and explore the relations between the clusters and earthquake properties. We use the dynamic time warping (DTW) distance and hierarchical clustering to group earthquakes based on the general shape of their STFs. We apply this to a global data base of 3529 STFs from earthquakes of magnitude greater than MW 5.5 between 1995 and 2018. The clusters exhibit different degrees of STF shape complexity, as measured by the number of prominent peaks. The clustering also indicates an association between the degree of STF complexity (or group) and earthquake source parameters. Thrust mechanism are in large proportion events with simple STF shapes and at all depths. In contrast, complex STF shapes correspond to shallow earthquakes with larger proportion of strike-slip mechanism and a generally longer duration. Moreover, we find that earthquakes with complex STF shapes tend to locate in the regions with complicated tectonics. These findings are corroborated by 2D dynamic modeling of earthquake ruptures on heterogeneous pre-stress and linear slip-weakening friction. We find a systematic variation of the simulated STF complexity with frictional properties. The clustering distribution provides useful constraints on elements of the frictional properties. In particular, the characteristic slip-weakening distance could be constrained to be over all short (< 0.1 m) and depth dependent.