In the thesis, we propose a method called the acceleration index (AI) calculate the variations of earthquake activities in Taiwan. We analyzed the influence of different factors(t0,t1,t2,∆tfor,∆tAfter) to the usual mean and recent mean, and used the distribution of AI map pixels to discuss the stabilize of mean value. Finally, the AI method is compared with the PI method. According to the results of the survey, the adjustment of the time t0 is helpful to realize the average value of occurrence of earthquakes in Taiwan. The relevant records of strong earthquakes can be eliminated based on the selection of the value of ∆tfor.