search

facebook / prophet

Tool for producing high quality forecasts for time series data that has multiple seasonality with linear or non-linear growth.
https://facebook.github.io/prophet
MIT License
18.47k stars 4.53k forks source link

Prophet with categorical or binary response (dependent) variable #2103

Open FlorinAndrei opened 2 years ago

FlorinAndrei commented 2 years ago

Is there any way at all to get meaningful results out of Prophet when the response variable is categorical or even binary (0 or 1)?

Basically, I have an up-or-down test that runs periodically. The output of the test is either 0 or 1. I'm looking for ways to forecast the response.

tcuongd commented 2 years ago

The response distribution is assumed to be normal and that's coded into the stan model, so no way to change that unless you make your own fork of the package and re-compile the model. Ben explored non-gaussian likelihoods for count data here: https://github.com/facebook/prophet/issues/1500 but I don't think we've explored binary responses specifically.

I haven't used it myself, but the tscount package in R might be a better fit for your task.