Search code examples
pythontime-seriespredictionarima

MA model is always giving same output in time series(python)


Does MA model can be used for prediction? In time series analysis, my ARIMA order is (0,1,7),which indicates that it is a MA model, unfortunately I am always getting a straight line for my prediction. Only first 7 values are different, as the lag is 7 here. And after that it predicts the same value.

my question is " is it normal that the model is predicting like that or there is a solution to overcome this problem. I would be grateful if anyone could help.


Solution

  • Yes, for a simple MA model it's normal to have a straight line for the prediction. Simple AR and MA models are intended to perform one-step-ahead predictions. In your case, you have a MA order of 7, so after your first seven "step ahead" the predictions start to repeat since you don't have new data.

    One way to avoid this would be to treat the predictions as ground-truth values and retrain the model with them. But don't expect this to improve your results significantly.

    For a more detailed explanation, I suggest reading this section from the Forecast: Principles And Practices by Hyndman.