Adapted Flower Pollination Algorithm for Lecturer-Class Assignment

摘要

Timetabling is considered a well-known combinatorial optimization problem that comes with challenges of assigning events or activities to resources such as time slots, space, and personnel that require these constraints are met. The most common timetabling problem is the constrained combinatorial University Course Timetable Scheduling (UCTS). The original Flower Pollination Algorithm (FPA) has limitations on its applications on combinatorial optimization problems like timetabling because it was originally designed to solve continuous optimization problem. This paper proposes an Adapted Flower Algorithm (AFPA) for lecturer-course assignment at the department level. In order to apply the algorithm to solve such problem, we introduce an improvement of the exploration and exploitation techniques. By applying in-memory randomization technique, this anticipates lessening the execution time and waste of assignments in searching and/or exploration phases. Genetic Algorithm (GA) operators are utilized to provide an effective distribution scaling factor in the pollination process. The experimental results with an artificial dataset with different constraints and complexity in the course assignments and preferences show that AFPA outperforms the traditional GA in terms of fitness value, convergence time. It also gives near global optimum solution and better results in terms of increasing or decreasing the complexity factor in the course preference list compared to GA.