time-series - 如何将二阶差转换回原始值？

Question

population2['value_first_order'] = population2['value'].diff()
population2['value_second_order'] = population2['value_first_order'].diff()

X = population2['value_second_order'].dropna()
x_train x_test = X[0:651], X[651:]

model = AutoReg(x_train, lags = 2).fit()
predictions = model.predict(start = len(x_train), end = len(X) - 1, dynamic = False)
predictions

如何将预测转换回原始值？

标签： time-seriesforecastingautoregressive-models