The present invention relates to a CAN data-based driver comprehensive quality evaluation method. The method comprises a first step of acquiring a latitude and a longitude, an engine rotation speed, a vehicle speed, a steer wheel rotation angle, an oil consumption, a throttle angle, and a brake pedal state, and calculating instantaneous oil consumption, acceleration, and a ratio of the engine rotation speed to the speed; a second step of determining evaluation factors, including a steering wheel rotation angle entropy H (theta), a steering wheel rotation angle rate V<theta>, a motor vehicle speed entropy H (v), an acceleration absolute value a, an acceleration strength V, a positive acceleration a, and a ratio phi of the engine rotation speed to themotor vehicle speed; a third step of establishing weight alpha of a single factor in evaluation of the driver comprehensive quality by using a vague layer analysis method, and establishing a desired weight vector Q; a fourth step of establishing as few as membership functions, and resolving a single factor vague determining matrix R; and a fifth step of resolving a determining vector theta=Q*R, and analyzing and resolving to obtain a determining set for evaluating the driver comprehensive quality. The method combines quantitative algorithm and qualitative algorithm, so as to evaluate the driver comprehensive quality and improve the evaluation accuracy.