A study of multiple-objective nurse rostering

• Main Authors: Tsai Tzung-Ming, 蔡宗明
• Other Authors: 李祥林
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/20065711254234761606

碩士 === 國立屏東科技大學 === 工業管理系 === 92 === The optimal nurse rostering problems are commonly formulated by goal-programming models with binary decision variables. Due to the combinatorial characteristics, optimal nurse rostering problems are classified as NP-complete and are difficult to solve when the sizes become large. Therefore, heuristics are applied to obtain better feasible solutions of the nurse rostering problem. All these heuristics have a common theme while searching the solution space, they all randomly select a new solution, hoping the search procedure can jump out of local optimum to often more promising search direction. Such selection may easily end up with an infeasible solution when there are constraints in the model, and is difficult to reach a feasible solution. Thus, the computational efficiency is suffered. This thesis constructs a goal programming model to obtain the optimal nurse roster. Simulated annealing algorithm is used to solve the model. In order to improve the computational efficiency, the model is further relaxed by adding penalty values to those infeasible solutions, resulting a model with only soft constraints. Thus, the solutions are all feasible. The results show that less then 1% of the solutions encountered are feasible, and SA can work very well with the relaxed model to obtain the final solution in less then 10 minutes. A comparison study of SA and LINGO is also conducted. It is found that LINGO requires longer computational time to obtain the final solution.