2025/08/31 by Tom Bensky, Bensky, Tom, K. G. Saunders +1
Computer Science · Decision Sciences · #Educational Technology and Assessment #FOS: Computer and information sciences #Intelligent Tutoring Systems and Adaptive Learning #Neural and Evolutionary Computing (cs.NE) #Scheduling and Timetabling Solutions
paper · pdf · doi:10.48550/arxiv.2509.06981
openalex publication_date 2025/08/31 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Here a genetic algorithm (GA) is presented that creates a teaching schedule for a university physics department by algorithmically assigning ∼200 classes to ∼50 professors for each of three academic terms per year. The algorithm is driven by chromosomes of the GA that encode proposed pairings between enumerated lists of professors and classes. The fitness of the pairings is measured by considering both contractual work constraints and individual teaching preferences. The algorithm uses standard crossover and mutation operations to seek ever more optimal schedules over many generations. Here we detail the implementation and performance of the algorithm, including some interpretability findings. Overall, we are very pleased with the algorithm, as it is typically able to converge within minutes, with over 90% of needed classes assigned. A metric is used to assign each professor's schedule a score, which measures how well their preferences were satisfied. These scores can be used to ensure longitudinal equity in the assignment of classes among professors.