|
David, very nice work. I have visited your web site
and checked out the information there. Your approach, which I would categorize
as letting humans handle the political concerns, may be the only viable
approach, I don't know. It is a common approach in many scheduling problems
(people don't like being pushed around by a computer?); even simple group
scheduling products such as outlook find it easier and safer to assist you,
letting you know common free times of a group rather than forcing you in to a
slot (every one is always free Friday afternoon). Obviously a computer program
can use things like the conflict matrix to help determine optimum placement of
course and do it faster an more accurately than a human.
The holy grail of scheduling however
remains:
A couple of questions about your loader. I notice a
search depth field, what is a common value for this? How deep do you suggest a
school go?
In my own tests, which would of course be greatly
affected by the priority system, (I don't allow humans to set scheduling
priorities, they do a bad job) I have never seen a student scheduled past about
20,000 tries (in 20 years my schedulers have made at least 100,000 runs),
although I current let the program try up to 100,000 (I think, don't have the
source code handy) (once I have the data from the database, the run is only 1-3
seconds, so I'm not much concerned about waste). It should be noted however that
most rejected students have an irresolvable conflict within a couple of hundred
try's.
I also notice a switch for seat stealing. I'm not
sure what you mean by this. It sound from the PDF on scheduling that you take
seats from students that have incomplete schedules, rather than trying to
reschedule an already scheduled student. In my scheduler I don't give rejected
students any seats, there is no point to it, so I'm really not stealing
seats.
If you are actually taking seats from scheduled
students (rescheduling them of course) I would like to know what is the expected
effect of this on the run. I.a. say in a typical large high school of 2000
students, how many more students would be schedule with this option vs. with out
it?
My own tests, years ago, showed that it was a lot
of extra work with little to show for it.
Once again very impressive work.
Jim
|