This Is AuburnElectronic Theses and Dissertations

Show simple item record

Staff Scheduling in Service Systems with Non-Stationary Arrival Processes


Metadata FieldValueLanguage
dc.contributor.advisorSmith, Jeffrey
dc.contributor.authorSamira, Shirzaei
dc.date.accessioned2021-12-06T04:43:40Z
dc.date.available2021-12-06T04:43:40Z
dc.date.issued2021-12-05
dc.identifier.urihttps://etd.auburn.edu//handle/10415/8041
dc.description.abstractThere are significant issues that cause customers’ dissatisfaction with service systems. One of the most critical frustrations deals with the unpredictable waiting time in the queue before getting service. This issue is even more noticeable in the systems with time-varying arrival rates. Variation in arrival rates demands appropriate staffing level responses. Understaffing leads to higher waiting times in the line, and over-staffing causes higher system costs. This dissertation covers different methods used to solve staff scheduling in this particular type of service system. The first approach is to discretize the time horizon into adjacent blocks. These periods are stationary queuing systems, and we can apply queuing methods to solve these problems. Next, we apply simulation-based optimization to find the staffing level appropriate for the system. Finally, we can use a staff scheduling algorithm to determine each staff member’s start and end working time. In the second method, we apply Deep Reinforcement Learning (DRL) methodology to perform staff scheduling. This method has many advantages over the previous one. Firstly, there is no need to estimate the location of change-points in the arrival process since the neural network can learn when and how to change the staffing levels. Next, unlike other approaches that require determining staffing levels first and then require carrying out staff scheduling, staff scheduling can be completed in one step by applying the DRL approach. Lastly, since all studies in the service systems area require data for analysis of their performance, the final objective of this dissertation is to model non-stationary processes by characterizing the arrival process and using the resulting models to generate the data with the same properties with matching the dispersion ratio. We will develop data generation algorithms when the arrival data are independent (renewal). The finding of this work and its future extensions can potentially help service systems like airports and call centers to improve their performance and increase their customers’ satisfaction.en_US
dc.subjectIndustrial and Systems Engineeringen_US
dc.titleStaff Scheduling in Service Systems with Non-Stationary Arrival Processesen_US
dc.typePhD Dissertationen_US
dc.embargo.statusNOT_EMBARGOEDen_US
dc.embargo.enddate2021-12-05en_US

Files in this item

Show simple item record