I need to predict tourist arrivals by country of origin, by segment (hotels, camping etc...), by purpouse of travel (leiser or business)... so by breakdown within breakdowns.

My question is; I do not know if it is better to start from the top and go down level by level or to forecast each segment and add up for the next level. So in the first case, I would first forecast the total tourist arrivals, then the part for each country, then for each segment for each country and so on... in the second case, I would forecast a specific segment (i.e. tourist arrivals from Germany in hotels for leisure purpose) and add up progressively.

How to know which of these methods would be more accurate?

Thank you very much!


Generally, this depends on the series. You could try both and see which works best.

There are also hybrid methods, see for example the work by "Hyndman et al" and their R package hts. Their methods involve modelling the series at all aggregation levels and modifying these preliminary forecasts to make them coherent.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.