Abstract: Software aging is a common phenomenon in a system that needs long-term operation. The existing analysis methods based on time series analysis mainly focus on autoregressive moving average (ARMA) models, not fully considered the seasonality or non-stationarity of the key indicators about software aging. This paper proposes a new software aging evaluation method based on seasonal autoregressive integrated moving average (ARIMA) model. The experimental results show that the method can well fit the software aging trend of seasonal load systems, and can achieve accurate prediction for supporting software rejuvenation.