A genetic algorithm for final exam scheduling of Isik University

Abstract

Universitelerde sıkca karsılaslan cizelgeleme problemlerden biri sınav haftası program nın ayarlanmasıdr. Sınav saatleri ayarlanırken zorunlu ve zorunlu olmayan kısıtlar goz onunde bulundurulur. Zorunlu olmayan kısıtlardan biri, aynı gunde istenilenden daha fazla sınava girecek olan ogrenci sayısının olabildigince az olması kısıtıdır. Isık Universitesinde bir gunde ikiden fazla finali olan ogrenciler mazeret sınavna girebilme hakkına sahiptir. Isık Universitesinde final programı ayarlanırken dikkat edilen diger kurallar icin tamsayılı programlama ile cozum uretilebilirken bahsedilen kısıt problemi bu yontemle cozulemez hale getirdi. Bu nedenle yaklasık bir yontem ile cozum uretme zorunlulugu dogdu. Gelistirilen genetik algoritmayı test ederken dort donemin verisinden faydalandk. Ayrca yeni yontem Java dilinde kodlanarak Isık Universitesinde kullanlan sınav cizelgeleme program finexa arayuzune entegre edildi.Anahtar sozcukler: Genetik algoritma, sınav cizelgelemesi, tamsayılı dogrusalprogramlama, memetik algoritma

Exam timetabling is a widely encountered scheduling problem at educational institutions. Typically, exam timetabling problems involve some hard constraints and several soft constraints that may vary from one institution to another. One of the soft constraints is that as few students as possible should have more than a predened number of exams on the same day. At Isk University, if students have more than two exams on the same day they are allowed to ask for makeup exams for the extra exams. While integer programming formulations with other constraints of Isk University could be solved to optimality via commercial solvers,incorporating the daily exam limitation rule proved to be intractable. Hence a genetic algorithm was developed. Using data from several semesters, numerical experiments were conducted to tune the developed genetic algorithm's parameters and test it. The new metaheuristic algorithm was also coded in Java programming language and integrated into nexa, the internally developed exam timetabling software at Isk University.Keywords: Genetic algorithm, exam timetabling, integer linear model, memeticalgorithm