Data-Driven Method for Constraint Customization in Optimization Models

Explore a 30-minute GERAD Research Center coffee talk on data-driven constraint customization in optimization models. Delve into Mahdis Bayani's research from Polytechnique Montréal, which addresses the challenge of adapting optimization software to end-user needs in various industries. Learn how machine learning techniques, particularly decision trees, can be used to extract implicit operational rules from previously implemented solutions and incorporate them into mixed integer linear programs (MILPs). Discover the extension of existing frameworks to accommodate non-linear and logical constraints, enhancing the customization capabilities of optimization models. Examine the practical applications of this approach through experiments conducted on knapsack and nurse rostering problems, demonstrating the value of data-driven constraint customization in real-world scenarios.

A data-driven method for constraint customization in optimization models, Mahdis Bayani