Multi-Objective Cost Function Optimization Using Artificial Bee Colony Algorithm With Enhanced Local Search for Course Scheduling Problem

In this study, local search ability of artificial bee colony algorithm (ABC) is improved for multi-objective cost function optimization and the proposed approach is applied to course scheduling problem. When considered in large scale, the course scheduling with lots of constraints turns into a problem that cannot be solved in polynomial time. This type of problems could only be solved by optimization algorithms. In this study, optimization of each objective function in multi-objective cost function is considered as a local solution and thus, the solution is assumed to be composed of a group of local solutions. The local search ability of bees is improved by sending more bees to worse local solutions of better solutions. According to this approach, two types of fitness functions, namely global and local fitness functions, are considered. Global fitness function is modeled as an increasing function as the mean-square error of local fitness values decreases, based on an approach in which bees pay more visits to poor local solutions. Thus, solutions having all local solutions synchronously better are prioritized throughout the solution space. Proposed approach is tested for course scheduling problem of Faculty of Engineering Departments in Mersin University. Experimental results show that proposed approach is more successful than standard ABC approach.