Evolutionary Approaches to the Partition/Timetabling Problem

Abstract

Recent research has yielded a variety of mainly evolutionary algorithm (EA) based methods for dealing with exam and course timetabling problems continually faced by educational institutions A problem as yet unaddressed in the timetabling research literature, however, is the partition/timetabling problem (PTP). This problem most typically arises early during a university term and requires the need to partition groups of students into small tutorial or lab groups, and then timetable these sessions. Even in cases where an institution has effectively ‘solved’ the main course or exam timetabling problems it faces, the continual need to address combined partition/timetabling problems still costs much staff time and effort. This article describes recent work which addresses the PTP. Three different techniques are tried, and results using these techniques axe compared on some real world PTPs. Best results so far are achieved by a two-stage method in which specially developed PTP heuristics are first used to derive a good partition of students into the several tutorial/lab (etc) sessions, and then a relatively standard timetabling EA generates a timetable based on this partition, incorporating operators which may slightly alter the partition.