Aiming at the fact that the mechanical vibration signal would exactly display non-stationary characteristics,a new method named as multiscale base-scale entropy（ MBSE） is introduced for measuring the complexity of time series,while MBSE which is based on base-scale entropy（ BSE） is defined to measure the complexity and irregularity of time series in different scale factors,and contains more and deeper information of time series. The concepts of BSE and MBSE are introduced firstly in this paper,and then they are used to measure the complexity of rolling bearing fault vibration signals. Finally,a new rolling bearing fault diagnosis method based on multiscale base-scale entropy and relevance vector machine is put forward. The experiment results show that the proposed method can diagnosis different fault state of rolling bearing effectively.