I am using statsmodels ARIMA (1,2,1) to predict the monthly demand for a product. The predictions look like they are shifted to the right by one month. I wonder if the statsmodels.ARIMA.Residuals.predict returns something different than the monthly demand predictions (maybe differences). Or is there anything else I might be doing wrong? I have also attached the acf and pacf plots. Thanks
Your results might not be wrong actually: Many predictions (about 50%) seem to be correct.
ARIMA works well in many cases, but the memory is limited. Maybe you should try to predict relative values instead of absolute ones.
I mean by relative values the values that increase (+20) or decrease (-30) over time. If your model learns on absolute value, it will recognize absolute patterns, and it might not be the case.
So, could you try to see if the results are better with relative values instead?