Logistics optimization w/ PyMC3?

Questions
1

I’m curious if anyone come across a resource oriented around PyMC3 for logistics, supply chain, inventory, …, optimization?

Most resources I stumble on take a deterministic approach (linear programming.) You can supply point estimates for each constraint, however, it’s not at all obvious how to approximate a global optimal configuration of parameters given simulated/sampled constraints.

For example, if you’re optimizing the number of items on hand in a warehouse, you might take into account the price an average customer is willing to pay (demand) and the number of units sold in a given time interval. To make use of linear programming, you’d need to use MLE estimates (or the means of a sampler outputs.)

More of an open ended question on- who is actively using PyMC3 for some form of logistics/supply chain analytics and optimization?

2

@twiecki has a blog article on this that you might find interesting: Using Bayesian Decision Making to Optimize Supply Chains — While My MCMC Gently Samples

2 Likes
3

@twiecki Hi Thomas, the link does not work anymore. Do you have a working version somewhere?
Thanks!

4

The link works for me

2 Likes
5

Thomas has been updating his blog recently, so maybe you hit it as he was in the middle of updating?

1 Like
6

Oh, yeah, it works for me now again.
It did not earlier today, but everything’s good now, thanks!

1 Like
7

The blog post works well for demand supply optimization over pricing. There’ a 1000 other things Bayesian modeling can do better than other types of statistics particularly in logistics, from inventory minimization, lead time estimation in multi level BOM/supplier scenario, and informing make vs buy decisions.

Hierarchical and overall flexible modeling make it all possible. Especially PyMC which simplifies specification and estimation. Here’s some more content for you to dig into
https://bayesiancomputationbook.com/markdown/chp_09.html

1 Like