Genetic Algorithms With Guided and Local Search Strategies for University Course Timetabling

Abstract

The university course timetabling problem (UCTP) is a combinatorial optimization problem, in which a set of events has to be scheduled into time slots and located into suitable rooms. The design of course timetables for academic institutions is a very difficult task because it is an NP-hard problem. This paper investigates genetic algorithms (GAs) with a guided search strategy and local search (LS) techniques for the UCTP. The guided search strategy is used to create offspring into the population based on a data structure that stores information extracted from good individuals of previous generations. The LS techniques use their exploitive search ability to improve the search efficiency of the proposed GAs and the quality of individuals. The proposed GAs are tested on two sets of benchmark problems in comparison with a set of state-ofthe-art methods from the literature. The experimental results show that the proposed GAs are able to produce promising results for the UCTP. Index Terms-Genetic algorithm (GA), guided search, local search (LS), university course timetabling problem (UCTP). I. INTRODUCTION T IMETABLING is one of the common scheduling problems, which can be described as the allocation of resources for tasks under predefined constraints so that it maximizes the possibility of allocation or minimizes the violation of constraints [40]. Timetabling problems are often made complicated by the details of a particular timetabling task. A general algorithm approach to one problem may turn out to be incapable for another problem, because certain special constraints are required in a particular instance of that problem. In the university course timetabling problem (UCTP), events (subjects and courses) have to be allocated into a number of time slots and rooms while satisfying various constraints. It is very difficult to find a general and effective solution for timetabling due to the diversity of the problem, the variance of constraints, and particular requirements from university to university according to the characteristics. There is no known deterministic polynomial time algorithm for the UCTP, since it is an NP-hard combinatorial optimization problem [18].

A hybrid genetic algorithm with local search and tabu search approaches for solving the post enrolment based course timetabling problem: Outperforming guided search genetic algorithm

“…(Please see [14] for more detail of these matrices.) The main objective of this problem is to minimize the number of soft constraint violations in a feasible solution, and infeasible solution, timetables which violate hard constraints, are always less importance than feasible timetable regardless of the number of soft constraint violations [13].…”

Interactive differential evolution for user-oriented image retrieval system

“…However, they generally suffer from excessively slow convergence to locate a precise enough solution efficiently due to lack of individual learning capabilities. Therefore, a lot of efforts have been made to improve the local search ability (García-Martínez and Lozano 2008), or to combine global and local strategies (Yang and Jat 2011). One remarkable work in this domain is Memetic Algorithm (MA) (Neri and Cotta 2012).…”

Optimal Sensor Placement for a Constellation of Multistatic Narrowband Pixelated Sensors

“…We assumed receivers placed at (0, 0), (10,0), and (20, 0), and computed (27) using the parameters given earlier for possible locations of a transmitter. puted using the recursive relationship of (25). Next, the PCRLB as a function of time is given by (26).…”

“…The surface is not convex; therefore, conventional optimization techniques which exploit convexity for efficient search through the solution space are not applicable. Techniques for global optimization of nonconvex multidimensional problems, including deterministic approaches such as branch-and-bound and heuristics-like decomposition [3], evolutionary algorithms [25], [29], memetic algorithms [27], [28], swarm methods [26], and stochastic methods [30], have received increased interest recently. In earlier work [21], we used an iterative-greedy heuristic with a related metric.…”