Multi-Objective Physician Scheduling using Native Binary Particle Swarm Optimization and Its Variance

Abstract

Physician Scheduling is one of the crucial operational activities in hospitals. The requirements of physician schedules involve numbers of physicians of different types, constraints from the physicians themselves and multiple regulations from the government and the hospital. Those requirements contribute to the complexity of physician scheduling model which makes it interesting to experiment and to solve. There have been many solutions proposed by researchers, particularly using metaheuristic algorithms, to automate physicians scheduling. Learning from the existing research results, this research experiments two metaheuristic algorithms which are considered simple, robust and effective to solve physician scheduling problem, i.e. Native Binary Particle Swarm Optimization (Native BPSO) with Sigmoid Transformation and its variance. The variation is done by removing particles’ local best positions from the velocity equation, with the intention to allow particles to move quickly towards global best position. The experiments were started from conducting the literature review of metaheuristic algorithms, followed by collecting physician schedules data from the selected hospital, designing the mathematical model of physician scheduling, developing and modifying the algorithms, testing and evaluating the results and at last, concluding the outcomes. Outcomes of the experiments comprise of three items, i.e. the mathematical model, the recommended algorithm and the best parameters’ values to be applied for the selected algorithm. Based on the experiments, the Native BPSO variant turns out to produce better result in terms of its fitness function and the number of days assigned for each physician in the schedule. Considering that the model, the algorithm and the parameters’ values have been implemented in a web-based application, it is ready for use by the selected hospital.