Thank you for your feedback. Since the Poisson Regression model is not a time series model, you cannot predict the future in the same way as you can generate predictions with a time series model such as the ARIMA model.

After you have trained a Poisson model, you can use the trained Poisson model to predict the 'outcome' if you know the regression variable values. In other words, you can predict y give X. In statsmodels, you can do this using the get_prediction() method on GLMResults. https://www.statsmodels.org/stable/generated/statsmodels.genmod.generalized_linear_model.GLMResults.get_prediction.html

Recollect from the article that you will get the GLMResults object when you call fit() on your Poisson model.

Does that answer your question?

'best