Thesis

A data-driven framework using AI predictive models to evaluate charterer satisfaction in the LNG shipping industry

Downloadable Content

Download PDF
Creator
Rights statement
Awarding institution
  • University of Strathclyde
Date of award
  • 2026
Thesis identifier
  • T18134
Person Identifier (Local)
  • 201960184
Qualification Level
Qualification Name
Department, School or Faculty
Abstract
  • The thesis addresses the need for a structured, data-driven approach to evaluating charterer satisfaction in the LNG shipping industry and so it introduces a framework called the Charterer Satisfaction Index (CSI). The CSI incorporates vessel performance predicting models that account for fuel consumption during laden passages, heel amount (fuel consumption in ballast condition), estimated discharging quantity, and the vessel’s emission footprint for each voyage. These models are critical to charterers’ decision-making in optimizing logistics efficiency. The charterers’ scope of work involves making decisions to match cargo requirements, originating from commodity traders who arrange cargo deals, with suitable vessels for cargo shipment (offered by shipping companies). By applying the results of predictive models, charterers can increase their income, while shipping companies can enhance their reputation. The CSI measures the deviation between predicted and actual performance values for each model, where smaller deviations indicate higher levels of satisfaction, reflecting more efficient decision-making by charterers, while larger deviations indicate lower satisfaction, indicating that the matching process is not optimized. To achieve this, a quantitative methodology is applied, supported by predictive machine learning modeling and high-frequency AI-driven data analysis. Seven machine learning models, Linear Regression, Lasso Regression, Gradient Boosting Regressor, AdaBoost Regression, XGBoost, Random Forest, Histogram-Based Gradient Boosting Regressor (HGBR), Light GBM using Root Mean Squared Error (RMSE), Coefficient of Determination (R²), and Mean Absolute Error (MAE). Among these, the Random Forest model consistently emerges as the best performing, achieving R² values ranging from 0.985 to 0.997 across three case studies conducted on three LNG vessels under both laden and ballast conditions. The case studies, based on two years of operational voyage data, demonstrate the robustness and predictive accuracy of the framework. Furthermore, expert interviews with industry professionals significantly contribute to the research, as their input guides both the selection of models for evaluating charterer satisfaction in terms of logistics efficiency and the weighting of each model’s relative importance within that assessment. The CSI is properly structured to capture and evaluate charterer satisfaction in a practical, realistic, and industry relevant manner.
Advisor / supervisor
  • Meer, R. van der (Robert)
  • Lazakis, Iraklis
Resource Type
DOI

Relations

Items

Thumbnail Title Date Uploaded Visibility Actions
file details PDF of thesis T18134 2026-09-23 Public Download