Enhanced Hospital Resource Management using Anticipatory Policies in Online Dynamic Multi-objective Optimization

摘要

Optimization problems in real-world settings are often dynamic and need to be solved online, i.e. as time goes by. Moreover, current decisions may have future consequences, which is called time-dependence [1], requiring anticipation of future situations to make well-informed decisions [1, 3].In real-world optimization problems, often more than one objective needs to be optimized at the same time. The optimum then is no longer a single solution but a set of solutions, i.e. the Pareto front. Population-based methods such as EAs are well-suited for multi-objective (MO) optimization [4].Earlier work on dynamic MO optimization is limited and focuses mostly on first definitions and algorithms, see e.g. [4]. No literature is published on performing anticipation in the optimization of multiple dynamically-changing objectives yet.We show how, for the dynamic MO problem of hospital resource management, anticipatory solutions can be obtained, providing better results than non-anticipatory solutions.