University of British Columbia
Integration of machine learning prediction into transportation planning of forest-based biomass
Abstract
dc:descriptionTransportation is a significant component of the forest-based biomass supply chain, accounting for more than one-third of the total cost. However, travel time delays further increase transportation costs, disrupt demand fulfillment, and reduce customer satisfaction, making it a critical challenge in logistics planning. Machine learning (ML) methods have been employed to predict potential delays and travel times to optimize logistics operations such as routing and scheduling. However, current applications of ML techniques for truck travel time estimation remain limited and have not yet been adopted in the forest industry. Furthermore, existing studies often emphasize prediction accuracy without integrating travel time predictions into practical routing and scheduling applications. In this research, a two-phase framework, combining ML-based travel time prediction with heuristic optimization-based routing, is developed. Using GPS data alongside influencing factors spanning weather conditions, road characteristics, and truck characteristics, the key factors affecting travel time and delay occurrence are identified (Objective 1). Multiple ML models are evaluated and compared based on prediction accuracy and computational efficiency to select the best-performing model (Objective 2). The resulting travel time predictions are then integrated into a heuristic optimization model to generate routing plans, enabling more robust logistics planning (Objective 3). Results showed that distance was the dominant influencing factor for travel time prediction, followed by road condition attributes such as proportion of uphill segments. For delay classification, road condition was the most important factor, such as road type and mean curvature. Among the evaluated ML models, XGBoost achieved the best balance of prediction accuracy (average error of 5.45 minutes) and computational efficiency (2.2 seconds). The predictions were subsequently integrated into the heuristic routing model, which reduced total travel time by up to 10% compared with those from the original dispatcher plans, with a computation time of only 3–5 seconds. These efficiency gains could result in an estimated cost savings of approximately $869,728 annually and a reduction in greenhouse gas emissions of approximately 384 tonnes of CO₂e per year. The results demonstrated that integrating ML-based prediction with optimization provided an effective decision-support tool for improving transportation efficiency and sustainability in forest-based biomass supply chains.
Degree
thesis:*- Name thesis:degree_name
- Master of Science - MSc
- Level thesis:degree_level
- master's
- Discipline thesis:degree_discipline
- Forestry
- Grantor dc:publisher
- University of British Columbia
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Wang, Aijun
- Contributors dc:contributor
-
- Sowlati, Taraneh
Rights
dc:rights- Statement dc:rights
-
- Attribution-NonCommercial-NoDerivatives 4.0 International
- Language dc:language
- eng
Identifiers
dc:identifier.*- Handle dc:identifier
- http://hdl.handle.net/2429/95429