Automated Lectures-Based Timetabling Generation Using Evolutionary Algorithm

Abstract

A timetable management system is designed and created to handle as much course data as fed while ensuring the avoidance of redundancy. Every school year, institutions of education face the rigorous task of drawing up timetables that satisfies the various courses offered by the different department. The difficulty is due to the great complexity of the construction of timetables for lectures, due to the scheduling size of the lectures, the high number of constraints and criteria of allocation, usually circumvented with the use of little strict heuristics, based on solutions from previous years. Also, the former timetabling systems did not consider the requests of lecturers as par the time convenient to fix their classes. This work employed Genetic Algorithm to generate timetable for faculty of agriculture courses. The hard, soft and float constraints for the system were formulated. The float constraint was included in hard constraints (system A) and then in soft constraints (system B). The repair strategy is also used for initializing a random population. The system was run with different parameters settings to obtain optimum results.
The results of the systems are: SA produced total of 4 soft constraints; SB produced a total of 9 soft constraints while SC produced 15 soft constraints. Thus, the accuracies of SA, SB and SC timetabling systems are 95.8%, 89.45 and 87.2%. This established that the lecturers request should be part of strong constraints for a conflict free timetabling.