Conceptual life cycle cost framework for marina pontoons: A predictive modelling approach based on Croatian marina data

DOI registering

Abstract

Effective marina asset management is constrained by limited cost transparency, fragmented maintenance records, and the absence of structured life cycle cost (LCC) decision-support tools specific to marina infrastructure. This study develops a conceptual LCC framework for marina pontoons based on predictive modelling and historical cost data from Croatian marinas. The proposed framework is designed to support early-stage design evaluation, maintenance planning, and long-term investment decision-making. The study is based on empirical data on construction, operation, maintenance, and end-of-life costs collected from 16 marinas along the Croatian coast over the period 2008-2018. A structured cost database was developed to support model formulation and validation. Multiple supervised machine learning algorithms were tested to evaluate their suitability for operational and life cycle cost estimation. Model performance was assessed using mean absolute error (MAE), cross-validation, and out-of-sample testing. The results show that Random Forest provides the most reliable performance for operational cost estimation, while Support Vector Machine achieves the best predictive performance for life cycle cost estimation. The proposed framework enables quantitative comparison of alternative design and maintenance scenarios and improves cost-informed decision-making in marina planning and management. The study contributes a structured cost database, a predictive LCC modelling framework, and a practical decision-support approach for marina infrastructure in data-constrained environments.