GENERATING DECISION SUPPORT INFORMATION FOR NURSE SCHEDULING INCLUDING EFFECTIVE MODIFICATIONS OF SOLUTIONS

When dealing with real-world problems, optimization models generally include only important structures and omit latent considerations that cannot be practically specified in advance. Therefore, it can be useful for optimization approaches to provide a “solution space” or “many solutions” containing a solution that the decision-maker is likely to accept. The nurse scheduling problem is an important problem in hospitals to maintain their quality of health care. Nowadays, given an instance, mathematical models can be applied to find optimal or near-optimal schedules within realistic computational times. However, even with the help of modern mathematical optimization systems, decision-makers must confirm the quality of obtained solutions and need to manually modify them into an acceptable form. Therefore, general optimization algorithms that provide insufficient information for effective modifications remain impractical for use in many hospitals in Japan. To improve this situation, we propose a method for a pattern-based formulation to generate information helpful in most practical cases in hospitals and other care facilities in Japan. This approach involves generating many optimal solutions and analyzing their features. Computational results show that the proposed approach provides useful information within a reasonable computational time.