0

This is my first question being asked here. I've thought about different methods to do it, but to no avail. I want to estimate a variable that is either 0 or a positive number. Then I want to use this estimation in a mixed integer linear programming (MILP) formulation. I have a big dataset with proper features that are labeled properly (the label is either 0 or a positive number). I'm thinking of using a machine learning approach to estimate this variable. It's important that the estimation is 0 (not a small number) when the label is 0. I was thinking of using a binary classification first (to divide zero and non-zero labels), and then a regression to estimate the positive numbers. But I'm not sure that I'll be able to model that in a MILP formulation efficiently. Any ideas? Can I use neural networks for this purpose?

I hope that I've been able to describe the problem well enough.

Mohammad Rajabdorri
  • 1
  • For MILP you might get better responses over at the [OR stack](https://or.stackexchange.com/). For something else, start with [zero-inflated regression](https://en.wikipedia.org/wiki/Zero-inflated_model). – Ben Reiniger Mar 23 '23 at 14:13
  • Thank you so much, Ben! Zero-inflated regression seems like an interesting solution. I will look into it. – Mohammad Rajabdorri Mar 24 '23 at 15:04

0 Answers0