A Non-worsening Memetic Algorithm for Constructing Feasible Timetables

Abstract

The choice of local search method in a Memetic Algorithm (MA) affects the effectiveness of the MA to produce feasible timetables. Since the local search itself depends on the definition of the neighborhood operator, we intend to understand the effect of the employed neighborhood operator to the performance of MA. In this work, we examine two typical neighborhood operators used in solving timetabling problems, that is neighborhood based on Move (reschedule lectures and replace rooms) and Swap (exchange the timeslot and room of two lectures). The results reveal the importance of defining neighborhood operator suitable with the problems. The performance of MA which surpasses the corresponding GA is also confirmed by the results which encourage the use of MAs in solving timetabling problems over GA.