Assigning proctors to exams using scatter search. (c2006)

Abstract

Scatter search is an evolutionary algorithm used to solve a multitude of real-life problems. It can be classified as a population-based methodology in the field of meta-heuristics. The main characteristic of scatter search is that it attempts to find improved solutions by combining others which are non-randomly generated. The scatter search methodology provides a general template consisting of five methods which can be fine-tuned to fit the specific problem under consideration. This flexibility has allowed the application of the scatter search to many combinatorial and optimization problems. One of the problems solved by the scatter search algorithm is the assignment of proctors to exams. This problem is encountered in every university at the time of final examinations when the scheduling of proctors is needed to oversee the students sitting for exams. This task has been reputed to be a very complex one because of the large number of exams and of the various constraints of this problem. Furthermore, when done manually, assigning proctors to exams is an extremely difficult and time consuming job for university staff responsible for this task and this entails that there is a need to automate the allocation of proctors to exams. The purpose of this project is to implement a computer-based decision support system that attempts to assign proctors to exams. The software was developed using Java, which is today's most advanced object-oriented language. The algorithm used is the scatter search methodology since it was found to provide good solutions for the problem under consideration. In addition, empirical results were obtained using real-life data and were found to give high-quality solutions that maximize the proctors' preferences while maintaining the fairness of the workload given to proctors.