0

In reaction to Sorry ARIMA, but I'm going Bayesian I was searching the Internet for a while but could not find any material regarding Bayesian ARIMA models in Python. Do you have any experience with this?

Septinel
  • 23
  • 4
  • 2
    There is [pymc](https://pymc-devs.github.io/pymc/), which is a general package for Bayesian modelling in python. Specifically for time series, there is [prophet](https://facebook.github.io/prophet/) by facebook, which I think also does something Bayesian under the hood. And there is also [TensorFlow probability](https://www.tensorflow.org/probability/) if you are used to TensorFlow. – sdcbr Dec 19 '18 at 14:14
  • Thank you! I'll surely take a look – Septinel Dec 19 '18 at 14:48
  • pystan is also a neat tool for Bayesian inference in python. Check out Chapter 10 of the 'Stan User Guide' for good examples. – nick Jan 06 '19 at 14:06
  • Did you have a further look at the options? Anything that stood out as particularly good? – nick Jan 08 '19 at 08:30
  • 1
    @nick Thanks! In my opinion the number one option is pymc3. Prophet looked really good at first but than I got a feeling that it is rather a black box. I was not able to adjust as I wanted. Finally I came across pydlm which is package for dynamic linear models which is preatty neat and easy to use. – Septinel Jan 08 '19 at 20:21
  • Thanks. Thats helpful to know :) – nick Jan 09 '19 at 09:28

0 Answers0