Enhanced Parallel-Particle Swarm Optimization (EP-PSO) Approach for Solving Nurse Rostering Problem

Abstract

A Nurse Rostering Problem is a highly-constrained combinatorial optimization problem, where we assign several nurses to shifts by violating minimum constraints. Due to massive number of constraints, these problems are difficult to handle manually. The advantage of automating the task is to generate a roster having not only high quality but also more flexibility by reducing the workload, time and effort of head nurses. The PSO algorithm is extremely dependent upon settings of control parameters and balance the exploration and exploitation in search space. These problems are avoided by proposed an ES-PSO Algorithm. Highly constrained problems have huge search space to find an optimal solution, hence to cover that space and find the solution in the stipulated time necessary to increase population. However, it may take more time for particle updating and fitness evaluations. To improve execution time of the compute-intensive task, we have used OpenMP and CUDA framework. The adapted algorithm improves the outcome by minimizing penalty and reduces stuff of compute-intensive tasks.