Journal of Intelligent Manufacturing, 15, 439±448, 2004 # 2004 Kluwer Academic Publishers. Manufactured in The Netherlands. The strategies and parameters of tabu search for job-shop scheduling FARUK GEYIK Industrial Engineering Department, University of Gaziantep, 27310 Gaziantep, Turkey E-mail: fgeyik@gantep.edu.tr ISMAIL HAKKI CEDIMOGLU Industrial Engineering Department, Sakarya University, 54040 EsentepeÐAdapazari, Turkey E-mail: cedim@sakarya.edu.tr Received February 2003 and accepted December 2003 This paper presents a tabu search approach for the job-shop scheduling problem. Although the problem is NP-hard, satisfactory solutions have been obtained recently by tabu search. However, tabu search has a problem-speci®c and parametric structure. Therefore, in the paper, we focussed on the tabu search strategies and parameters such as initial solution, neighborhood structure, tabu list, aspiration criterion, elite solutions list, intensi®cation, diversi®cation and the number of iteration. In order to compare some neighborhood strategies and tabu list length methods, a computational study is done on the benchmark problems. Keywords: Tabu search, neighborhood, tabu list, job-shop scheduling 1. Introduction Scheduling concerns the allocation of limited resources to tasks over time (Pinedo, 1995) and the ef®ciency of this allocation. The resources and tasks may be in many forms. In a manufacturing environ- ment, e.g., the resources may be machines and the tasks may be operations. Production scheduling together with production planning is an important function which determines the ef®ciency and produc- tivity of a manufacturing system. However, production scheduling is not an independent function; there are many elements which affect it. For instance, precedence constraints, due dates, production levels, lot-size restrictions, priority rules etc. In addition, the basic element which affects the scheduling is essentially process planning (Geyik and Cedimoglu, 1999). There are commonly two kinds of feasibility constraints in production scheduling problem: machine capacity and precedence constraints. A schedule is any feasible solution of these constraints (Baker, 1994). In principle the number of feasible schedules for any job-shop problem is in®nite because it can be put an arbitrary amount of idle time between successive operations. Even if no idle time is left, the number of possible semi-active schedules will be excessive, that is about n! m Ðwhere n is the number of jobs and m is the number of machines. In this case, to ®nd an optimum schedule must be consumed excessive computational timeÐeven it sometimes will be impossible. According to Pinedo (1995), if a scheduling problem does not have an ef®cientÐso- called polynomial timeÐalgorithm, it is called non- deterministic polynomial hard (NP-hard) problem. So, the job-shop scheduling problem (JSP) is NP-hard. Although the solution of scheduling problems optimally is dif®cult, the most ef®cient ways of exact solution methods are the branch and bound algorithm and dynamic programming. Some of works in these ®elds are Carlier and Pinson (1989), Applegate and Cook (1991), Brucker et al. (1994),