MSc Thesis Defense: Başak Çarhacıoğlu, A Machine Learning-Enhanced Tabu Search Approach for the Stochastic Vehicle Routing Problem with Flexible Deliveries and Soft Time Windows,Date & Time: June 24, 2026 – 10:30 AM, Place: FENS G032

A Machine Learning-Enhanced Tabu Search Approach for the Stochastic Vehicle Routing Problem with Flexible Deliveries and Soft Time Windows

Başak Çarhacıoğlu
Data Science, MSc Thesis, 2026

Thesis Jury

     Asst. Prof. Duygu Taş Küten (Thesis Advisor)

  Assoc. Prof. Öznur Taştan Okan

  Asst. Prof. Amine Gizem Tiniç

Date & Time: June 24th, 2026 –  10.30 AM

Place: FENS G032

Keywords : Vehicle Routing Problem, Flexible Deliveries, Stochastic Travel Times, Tabu Search, Machine Learning

Abstract

This thesis addresses the Vehicle Routing Problem with Flexible Deliveries and Stochastic Travel Times (VRPFD-STT), in which customers may specify multiple delivery locations associated with different time windows throughout the day and travel times are uncertain. The objective is to minimize a weighted combination of operational costs, including travel distance, vehicle usage, and expected overtime, together with service-related costs arising from expected time-window violations. To effectively solve the VRPFD-STT, this thesis proposes a tabu search-based solution approach enhanced with machine learning, where a logistic regression model is integrated into the neighborhood exploration process to prioritize promising moves. The algorithm also incorporates diversification and intensification mechanisms, route-pool-based solution reconstruction, and post-optimization. Relative to upper bounds available from the literature based on column generation, the algorithm yields an average gap of 2.35%. In addition, computational results show that, for the deterministic counterpart considering hard time windows, the proposed approach achieves 31 optimal solutions out of 60 benchmark instances, with an average optimality gap of 1.37%.