{"title":"Behind-the-Scenes Weight Tuning for applied nurse rostering","authors":"Elín Björk Böðvarsdóttir , Pieter Smet , Greet Vanden Berghe","doi":"10.1016/j.orhc.2020.100265","DOIUrl":null,"url":null,"abstract":"<div><p>Although researchers have developed countless nurse rostering algorithms throughout the years, the paradigm of manual scheduling continues to hinder their application in practice. While manual scheduling gives practitioners full control in assigning nurses to shifts based on their knowledge of the personnel, it has some severe drawbacks. Manual scheduling is tremendously time-consuming and often fails to reach organizational targets, as practitioners need to address numerous constraints and objectives, which frequently conflict with one another. Until now, most nurse rostering formulations have employed weighted sum objective functions that rely on manually-set weights. Understanding the impact of those weights, and thus selecting appropriate values for them, is not trivial. Consequently, the optimization objective often does not capture the desired outcome, resulting in poor quality rosters with an unacceptable combination of constraint violations. This paper introduces a general methodology, <em>Behind-the-Scenes Weight Tuning</em>, which uses measurable targets for guidance in order to automatically set weights. As the methodology does not require practitioners to provide accurate objective weights, the level of manual effort is substantially reduced. Outcome of experiments has shown that by enabling the computer to make quantitatively-supported decisions in this manner, we consistently obtain better rosters than when relying on practitioners to set appropriate weights.</p></div>","PeriodicalId":46320,"journal":{"name":"Operations Research for Health Care","volume":null,"pages":null},"PeriodicalIF":1.5000,"publicationDate":"2020-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1016/j.orhc.2020.100265","citationCount":"2","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Operations Research for Health Care","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S221169232030045X","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"HEALTH CARE SCIENCES & SERVICES","Score":null,"Total":0}
引用次数: 2
Abstract
Although researchers have developed countless nurse rostering algorithms throughout the years, the paradigm of manual scheduling continues to hinder their application in practice. While manual scheduling gives practitioners full control in assigning nurses to shifts based on their knowledge of the personnel, it has some severe drawbacks. Manual scheduling is tremendously time-consuming and often fails to reach organizational targets, as practitioners need to address numerous constraints and objectives, which frequently conflict with one another. Until now, most nurse rostering formulations have employed weighted sum objective functions that rely on manually-set weights. Understanding the impact of those weights, and thus selecting appropriate values for them, is not trivial. Consequently, the optimization objective often does not capture the desired outcome, resulting in poor quality rosters with an unacceptable combination of constraint violations. This paper introduces a general methodology, Behind-the-Scenes Weight Tuning, which uses measurable targets for guidance in order to automatically set weights. As the methodology does not require practitioners to provide accurate objective weights, the level of manual effort is substantially reduced. Outcome of experiments has shown that by enabling the computer to make quantitatively-supported decisions in this manner, we consistently obtain better rosters than when relying on practitioners to set appropriate weights.