EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
3599. Stochastic scheduling in the presence of side data - case: surgery planning
Invited abstract in session WA-6: Advancements of OR-analytics in statistics, machine learning and data science 16, stream Advancements of OR-analytics in statistics, machine learning and data science.
Wednesday, 8:30-10:00Room: 1013 (building: 202)
Authors (first author is the speaker)
| 1. | Ricardo Otero
|
| KU Leuven | |
| 2. | Erik Demeulemeester
|
| KBI, KU Leuven |
Abstract
In this research, we focus on the stochastic Advance Scheduling Problem with Downstream Capacity Constraints (ASPDCC) in the case where the decision maker has access to side data. We begin formulating the stochastic ASPDCC as a two-stage stochastic program using a sample average approximation (SAA) approach. Unlike the common method to characterize uncertainty, which assumes a lognormal distribution as the true data-generating process for each surgical specialty, we propose a data-driven framework that integrates a machine learning algorithm (decision trees) to estimate the conditional probability distribution for the surgery durations and the length of stay (LoS), in the presence of side data. Finally, to speed up the solution of the SAA model, we use a benders decomposition incorporated into a branch-and-cut framework. To justify the value of using side data, we perform extensive experimentation using real data from a reference hospital in Bogota, Colombia. Results show that our algorithm generates schedules that outperform the traditional method of fitting a lognormal distribution for each surgery specialty.
Keywords
- Analytics and Data Science
- Health Care
- Combinatorial Optimization
Status: accepted
Back to the list of papers