Enhancing data quality in maritime transportation: A practical method for imputing missing ship static data

Maritime transport research, including emissions estimation, shipping network design, and bunker management relies on complete high-quality data. Incomplete ship static data often lead to bias and misleading conclusions. Existing imputation methods either depend on small data samples with poor imputation accuracy or are overly complex with practical limitations. This study addresses these issues by evaluating existing methods and proposing a new imputation approach to enhance the quality of ship static data. First, we introduce a stepwise multiple nonlinear regression method to simplify the imputation process and improve accuracy. Second, we propose a novel evaluation metric, the coverage rate, to assess the model performance. Finally, from a total of 14 models, we use a decision matrix to select the optimal model for imputing missing ship static data. These models are applied to a real dataset with missing values and cross-validated using multiple databases to ensure robustness. The proposed method maximizes the coverage rate, approaching nearly 100 percent for missing data. The most significant improvement was observed in main engine RPM imputation, where the average adjusted R-squared increased by at least 20.74%. Based on a large training dataset of 38,018 ships, this method can be directly applied to other maritime transport studies. • Proposed stepwise nonlinear regression for ship static data imputation • Introduced coverage rate metric to assess successful data imputation. • Models tested on 38,018 ships and cross-validated for robustness. • Developed decision matrix to select optimal imputation model for various scenarios