Scheduling final examinations for a university with different requirem
ents is a tedious and laborious task. This paper describes an expert s
ystem called Examination Management System (EMS) which can complete th
e scheduling very efficiently In EMS we select the most difficult-to-s
chedule examination as a candidate for the next scheduling step. When
the scheduling process cannot proceed any further for lack of time or
room, the system releases the resources occupied by the easiest-to-sch
edule examinations (among the scheduled) and assigns them to the exami
nation currently under consideration Some statistical learning methods
are applied to dynamically change the focus among the considerations
for the requirements or constraints such as direct conflicts and room
capacity. Any requirements and expertise can be easily added into or r
emoved from the system, but some trivial ad hoc requirements are neede
d to be handled in the pre-scheduling part. EMS has been used successf
ully for two local universities with much rime saving and quality impr
ovement. (C) 1997 Elsevier Science Ltd.