This paper proposes a new memetic evolutionary algorithm to achieve explicitlearning in rule-based nurse rostering, which involves applying a set ofheuristic rules for each nurse's assignment. The main framework of thealgorithm is an estimation of distribution algorithm, in which an ant-minermethodology improves the individual solutions produced in each generation.Unlike our previous work (where learning is implicit), the learning in thememetic estimation of distribution algorithm is explicit, i.e. we are able toidentify building blocks directly. The overall approach learns by building aprobabilistic model, i.e. an estimation of the probability distribution ofindividual nurse-rule pairs that are used to construct schedules. The localsearch processor (i.e. the ant-miner) reinforces nurse-rule pairs that receivehigher rewards. A challenging real world nurse rostering problem is used as thetest problem. Computational results show that the proposed approach outperformsmost existing approaches. It is suggested that the learning methodologiessuggested in this paper may be applied to other scheduling problems whereschedules are built systematically according to specific rules
展开▼
