The invention discloses data driving-based bus load characteristic analysis. According to the data driving-based bus load characteristic analysis, clustering analysis is carried out on a daily bus load curve, wherein the waveform sizes of electric quantity at different time periods are fully considered; the tie periods with relatively large fluctuation are endowed with relatively high weights so that the degree of distinguishing the bus load through a clustering algorithm is enhanced; division is carried out on a bus according to the bus load curve clustering result; four indexes such as model switching entropy, relative fluctuation ratio, daily average load and temperature sensitivity are put forward on such basis, so as to describe the bus load characteristics from different dimensionalities; and finally, classification is carried out on the bus load by utilizing a K nearest algorithm according to extraction and characteristics, so as to finally carry out example simulation on Guangzhou actual bus load data. According to the data driving-based bus load characteristic analysis, the indexes can be used for well describing the bus load characteristics and obtaining relatively good classification effect.