A Generic Two-Phase Stochastic Variable Neighborhood Approach for Effectively Solving the Nurse Rostering Problem

Abstract

In this contribution, a generic two-phase stochastic variable neighborhood approach is applied to nurse rostering problems. The proposed algorithm is used for creating feasible and efficient nurse rosters for many different nurse rostering cases. In order to demonstrate the efficiency and generic applicability of the proposed approach, experiments with real-world input data coming from many different nurse rostering cases have been conducted. The nurse rostering instances used have significant differences in nature, structure, philosophy and the type of hard and soft constraints. Computational results show that the proposed algorithm performs better than six different existing approaches applied to the same nurse rostering input instances using the same evaluation criteria. In addition, in all cases, it manages to reach the best-known fitness achieved in the literature, and in one case, it manages to beat the best-known fitness achieved till now

A tensor based hyper-heuristic for nurse rostering

“…On the instance Valouxis-1, both algorithm can achieve the best known result (20), although much later than the state-of-the-art (Solos et al, 2013). Similarly, on Ikegami and ORTEC instances as well as BCV-3.46.1, the state-of-the-art performs better.…”

“…In (Solos et al, 2013), a generic two-phase variable neighbourhood approach was proposed for nurse rostering. After determining the value of the parameters which govern the performance of the algorithm, a random popu-lation of candidate solutions is generated.…”

Α two-phase adaptive variable neighborhood approach for nurse rostering

“…More precisely, regarding the approach presented in [11], which is a former attempt of ours to solve the nurse rostering problem, the proposed algorithm has the following major differences:…”

“…More precisely, it is a two-phase stochastic variable neighborhood approach, which uses nine different swap mechanisms. Some of them are innovative while the rest ones are well known and established swap mechanisms which, in our case, are applied in a total different way compared to other variable neighbourhood approaches presented in the literature [8,11,20,30]. More details concerning the differences of the proposed approach with former variable neighbourhood approaches are presented in Section 3.…”

“…They are applied to complex optimization problems in cases where other optimization methods have failed to be either effective or efficient. Heuristic and metaheuristic representative methods solving the nurse rostering problem include evolutionary algorithms [5,6], memetic algorithms [7], variable neighborhood search [8][9][10][11], scatter search [12,13], tabu search [14,15], iterated local search [16], simulated annealing [17], ant colony optimization [18], particle swarm optimization [19], etc.…”