Workforce Scheduling & Routing

ML classifier for algorithm selection in scheduling problems — 40% performance improvement

A gradient-boosting classifier that predicts the best optimization algorithm given a problem instance and runtime constraint, achieving 40% higher performance than any single algorithm applied uniformly. This work is the subject of a co-authored paper currently in review.

Key contributions:

  • Formulated algorithm selection as a classification problem over scheduling problem features
  • Trained gradient-boosting model (scikit-learn) on a diverse benchmark suite
  • Achieved 40% performance improvement over best single-algorithm baselines
  • Co-author paper submitted for peer review

Stack: Python, Scikit-learn, Pandas, NumPy

Publication: Manalp, C. K., Erol, A. K., Erol, K., & Evrendilek, C. (2024). A Holistic Approach to Workforce Scheduling and Routing.