Budi Santosa Andiek Sunarto Arief Rahman Indusrial Engineering Institut Teknologi Sepuluh Nopember Kampus ITS Sukolilo Surabaya, Indonesia phone +62315992364 email : budi_s@ie.its.ac.id ABSTRACT Airline crew rostering is the assignment problem of crew members to planned rotations/pairings for certain month. Airline companies have the monthly task of constructing personalized monthly schedules (roster) for crew members. This problem increased more and more complex and difficult while the aspirations/criterias grew to assess the quality of roster and the constraints increased excessively. This paper proposed the differential evolution (DE) method to solve the airline rostering problem. Different from the common DE, this paper presented random swap as mutation operator. The DE algorithm is proven able to find the near optimal solution accurately for optimization problem. Through numerical experiments with some real datasets, DE showed more competitive results than two other method, column generation and MOSI (the one used by the Airline). DE produced good results for small and medium datasets, but still showed reasonable results for large dataset. For large crew rostering problem, we proposed decomposition procedure to solve in more efficient manner using DE. Keywords differential evolution, crew scheduling, pairing, rostering. MSC code: 90-08, 90B 1. INTRODUCTION In airline industry, development of crews rostering plan which be able to produce the high utility of crews become the priority in human resources department. It is estimated that optimization software which have been developed for airline could save more than US $ 20 million per year [1]. Saving 1% in crew utilization can save cost largerly. Though airline crews scheduling became attention in many operation research literature such as ([1]; [2]; [11]; [13]; [30];[37]) but airline crews scheduling remains to become the main attention for many researchers that is caused by level of complexity and difficulty to solve it. Therefore method and approach which are used to solve it more and more grow to get the better result in optimality side and speed of computational time. Beside of reasons which mentioned previously. Generally, solving airline crew scheduling is done by decomposition approach ([5]; [25]), it devides problem to two problems, that are crew pairing and crew rostering. Crew pairing is done to get initial feasible solution, that is sequence of flight which begin and end on the same home base. Crew rostering assigns pairings which were arranged for the certain month to set of crews based on individual calender. Decomposition approach is very effective to solve the difficult and complex problem but this method loss the global treatment since crew pairing and crew rostering done separately. Some other researchers developed the integrated approach to overcome obstacle, such as Souai and Thegem [36], where crew pairing and rostering were done simultantly to get level of better optimality. Many optimization methods have been developed to solve crew scheduling to increase roster quality and to improve computational time such as simulated annealing [25], genetic algorithm [37], tree search algorithm [3], hybrid genetic algorithm [24] and GASA hybrid algorithm [41].