中文说明：
ARIMA 模型是在平稳的时间序列基础上建立起来的，因此时间序列的平稳性是建模的重要前提。检验时间序列模型平稳的方法一般采用 ADF 单位根检验模型去检验。当然如果时间序列不稳定，也可以通过一些操作去使得时间序列稳定（比如取对数，差分），然后进行 ARIMA 模型预测，得到稳定的时间序列的预测结果，然后对预测结果进行之前使序列稳定的操作的逆操作（取指数，差分的逆操作），就可以得到原始数据的预测结果。
English Description：
ARIMA model is built on the basis of stationary time series, so the stationarity of time series is an important prerequisite for modeling. ADF unit root test model is generally used to test the stationarity of time series model. Of course, if the time series is unstable, you can also use some operations to make the time series stable (such as taking logarithm and difference), and then perform Arima Model prediction is used to obtain the prediction results of stable time series. Then, the prediction results of the original data can be obtained by the inverse operation of the operation that makes the series stable before the prediction results (taking the index, the inverse operation of the difference).