Updated: 22 July 2007


Instructions by Liviu Lalescu


These are some small instructions which you have to follow in order to input a solvable data set.


If FET cannot find a good timetable for your school, make sure to write to the author, as FET needs to be improved. Maybe small changes to your datafile can bring good solutions, or maybe there are aspects which have to be changed in FET.


1) You can use for the constraint min n days the weight you want. It can be 100%, but please make sure your timetable is not too constrained. Please take care of the impossible constraints, they should have 0 weight (percentage). For instance, if a teacher teaches only two days per week and has 3 math lessons for a group, then it is clear that the constraint cannot be respected, so the correct way is to specify a 0 weight (percentage) for min n days constraints.


2) It is advisable to restrict the number of hours and days to the needed value, not larger. The reason is to guide FET to obtain a good solution. It is not as good to add a large number of hours per day and add constraint teachers/students max hours daily, as I observed in practice.


3) It is advisable to use not available, break and preferred time(s) to make restrictions. These will not slow down the generation.


4) If you have a course lesson with say 4 activities per week, difficult lessons, which you would like to schedule in the beginning of the day (say 3 out of 4 need to be in the first half of the day), you can add constraint activity preferred times for the first three activities, with preferred times in the first half of the day. This approach worked with the sample German high-school data.


5) A real case: for the sample German file, FET usually finds a solution fast. I tried the following: increased the day from 6 to 12 hours, and let the students begin hours at any time (removed students early). I kept the constraint teachers max gaps per week to 3. But to my surprise, even if the data was less constrained, FET failed to find a solution. The reason is that the teachers have a larger possible palette of options, and FET is not able anymore to optimize their timetable so not to have more than 3 gaps per week.


The advice is to help FET to have the data as a human timetabler would have it, helping him by not letting him with a very large palette of options for each activity and on the other hand constraining him with max gaps for instance. Try firstly to see if the timetable can be generated without teachers max gaps per week, and lower the max gaps limit afterwards.