This repository consists of the code I edited and developed which solves a unique timetabling problem of a large academic department. The code uses an evolutionary algorithm, a simulated hardening process and a type of general adversarial network which produces a range of valid timetables
Timetabling algorithms such as metaheuristics, mathematical programming, graph coloring, hyper heuristics have been studied and applied extensively in various fields such as healthcare, transportation systems and in educational institutions to solve complicated optimization problems like NP-hard. Concentrating on university course timetabling problems (UCTP), particularly for a specific university, there is a literature gap in the constraints and the unusual module teaching approach that WMG possesses in addition to a lack of literature in utilizing machine learning (ML) techniques to solve UCTP. Consequently, this piece of research, using a design science methodology and pragmatic approach, aims to employ a hybrid metaheuristic and ML algorithm to effectively automate much of the current manual and strenuous timetabling task for WMG administration where the main focus is on exploiting generative models like General Adversarial Networks to enhance the diversity and number valid timetabling solutions generated relative to only using Evolutionary or Genetic Algorithms.
This results in offering flexibility to university stakeholders by reducing much pressure and negotiations that are otherwise necessary in order to agree on teaching slots of instructors.