On the Development of MCMC Models

[psygo]

Since this is a very frequently requested feature, this issue was created to foment the development of MCMC (Markov Chain Monte Carlo Methods) implementations of the so far analytical models inside the lifetimes library.

If you can, please collaborate with (non-exhaustive list that will be edited in the future):

• References
  • Other libraries' implementations.
    • BTYDplus.
• Articles
• Pseudo or Actual Code
• Other Issues with Related Content
  • #168

[zdhiman]

pydata Seattle 2017: Implementing and Training Predictive Customer Lifetime Value Models in Python Found this talk insightful. Unfortunately the code has been recently removed from github.

Will try to reproduce and add a notebook soon! 👍

[romanoww]

@1dhiman please post when you do!

[sophia-wright-blue]

@1dhiman , is the notebook at this repo the one that you are referring to?

https://github.com/jrgauthier01/oreilly-intro-to-predictive-clv

[sophia-wright-blue]

@psygo , can you provide more details on the MCMC functionality that you'd like to add as a feature?

[ColtAllen]

@sophia-wright-blue I've been working on a new modeling backend based on the MCMC functionality of pymc in my fork for lifetimes, and I just now posted the PR:

https://github.com/ColtAllen/lifetimes/pull/24