| Summary: | In this thesis, the examination timetabling will be modelled for the University of Iceland. As the situation has been, the timetabling problem has been solved manually requiring 5-7 days each semester. By making models that solve the problem could not only lead to less time spent in timetabling but could potentially give a better solution for the students. A two-phase approach was chosen for the problem where Phase I schedules the examinations into slots while Phase II schedules the examinations to suitable rooms. The key objective of Phase I was to make a conflict free timetable but also minimize same day examinations for students, students facing two examinations within 24 hours and students not receiving one day off before an exam. Phase II had multiple objectives but the overall goal was to minimize the rooms and buildings being used for each exam. The results achieved by using the models gave promising results for both phases. The main drawback was the computational time for Phase I. Therefore, a low-level heuristic was proposed. The solution achieved by the heuristic approach gave slightly worse results but required less computational time. By using the heuristic approach, Phase I was split into two sub-phases but by splitting a problem into two sub-phases can lead to an infeasibility. Therefore, a new way must be proposed in the future using a curriculum based scheduling. Í þessari ritgerð verða tvö líkön gerð fyrir Háskóla Íslands (HÍ) sem leysa próftöflugerð háskólans. Á undanförnum árum hefur vandamálið verið leyst handvirkt og tekur alla jafna 5 − 7 vinnudaga að leysa hvert próftímabil. Með innleiðingu slíkra líkana gæti hlotist mikill ávinningur í mögulegum vinnusparnaði og ekki síst fyrir nemendur sem gætu fengið betri próftöflur. Tveggja fasa lausnaraðferð var beitt þar sem Fasi I raðaði prófum í lotur meðan Fasi II notaði lausn frá Fasa I og raðaði prófum í viðeigandi stofur. Við röðun prófa í lotur voru grunngildi röðunarinnar að hafa sem fæsta nema í prófi samdægurs, í röð (þeas innan 24 ...
|
|---|