但做了这么多数据变换,如何将预测值还原呢? You need to reverse all transformations. 发现对于做arima预测的序列,应避免趋势性和季节性,一般采用取对数、季节调整和差分的办法 That's basic skills for TS.
在arima模型预测过程中,对于序列平稳性问题,一直存在疑问。高铁梅书上没有给出详细的解释,只是说必须平稳. if you means stationary. The reason is simple. There must have some similarity of past to current data and future data., otherwise no way to predict. Do you think you can predict unpredictable things? The answer is No. You'd better read some books written in Englis. Bless