Authors:
Quoc Tien Au
and
Faiyaz Hasan
Affiliation:
GradeSlam, Montreal, Canada
Keyword(s):
Scheduling, Online Tutoring, Integer Linear Programming, Heuristics, Staff Scheduling, Rostering, Constraints, Objective Function, Adaptive Constraints, Constraint Relaxation, Robustness, Robust Scheduling.
Abstract:
Scheduling personnel is an important aspect of many organizations and can rapidly become unsustainable to perform manually. It is crucial that the automated algorithm consistently generates a high quality schedule. This paper discusses a linear integer program with a set of constraints, developed in collaboration with a scheduling expert, required to generate reliable schedules for a live online tutoring platform. The focus in the algorithm is shifted away from the optimization of an objective function to adopting hard constraints that guarantee any solution will be sufficiently good. Furthermore, a graceful degradation protocol to implement fuzzy boundary conditions for the constraints ensure that appropriate compromises can be made in most cases to find a solution. Lastly, a comparison of the automated algorithm to a domain expert shows a 5% higher topic coverage and 10% lower cost in favor of the former which has led to a rapid adoption of the automated scheduler by the organizati
on.
(More)