System identification by Auto-regressive Model that based on time series data is widely used for the purpose of estimating a mathematical model of system in various fields. However, this approach needs to estimate according to the individualized theory. In this study, considered whether this technique could be applied in the MEMS oscillator. The AR model method is generally used to estimate the unknown parameter that constitute system. So, MEMS oscillator system parameter changes with packaging before and after, can be expected to analyze. In addition, white noise including using a variety of input frequencies can be expected to establish high-throughput extraction parameters.