I used prophet for time series forecasting. the holiday effect for Christmas and thanksgiving were captured well. The data show strong positive effect for the Easter Sunday('lower_window': -2, 'upper_window': 1), but the the prediction doesn't capture effect well. See the plot below for your reference.

What might cause this issue? how to improve it? many thanks

  • $\begingroup$ stats.stackexchange.com/questions/417144/… discusses some of the weaknesses that makes prophet a less-than-perfect prophet. A primary weakness is the inability to detect pre and post effects around holidays among a number of other anthropogenic effects. For example in my work in retail, the buildup to Christmas starts many many days in advance of Christmas . Finding that kind of structure is one of the features of AUTOBOX , which I have helped to develop. If you post your data and the country of origin , I migh $\endgroup$ – IrishStat Mar 18 at 12:35
  • $\begingroup$ Could you clarify the sense in which the prediction "doesn't capture effect well"? How do you know that and how is that shown in your plot? $\endgroup$ – whuber Mar 18 at 12:40

Browse other questions tagged or ask your own question.