I have an existing given schedule for a hospital department. I use supervised learning from historical data to predict the reservation durations providing determined process times. A requirement is not to change the original schedule 'too much'. So I will include a cost in respect to time for distance from the original schedule. I believe a local search approach (Tabu, SA, GA) will work well, where I always start from the original schedule. Are there any other algorithms or approaches that may better suit this problem?
Hi, deterministic vs stochastic scheduling refers to the a priori knowledge, in my case knowledge of the process times. A deterministic scheduling problem implies that the process times and other information is given, as oppose to in a stochastic scheduling problem where the process times are stochastic. The first paragraph in this book explains it well. http://www.springer.com/mathematics/probability/book/978-90-277-1397-1
Deterministic scheduling problems are NP-hard (in some cases NP-complete).
For further reading I strongly recommend Micheal Pinedo's book on Scheduling.
Do heuristics like GA find the global optimal, or do they find an acceptable local optimal? I think this is what you mean?
- Badges
- Science topic
- Similar topics
- Computer Science
- Data Mining