Estimating Passenger Capacity of Ships with Linear Regression Models

Accurate prediction of passenger ship capacity is essential for ship design, fleet management, and planning. In this study, verified technical and operational data from about 2,000 passenger ships were analyzed. The dataset included key variables such as passenger capacity (P), gross tonnage (GT), draft (T), length (L), beam (B), deadweight tonnage (DWT), and main engine power (EP). First, the distributions and correlations of all these variables were examined. Variables that most strongly affected passenger capacity and reduced the risk of multicollinearity were carefully selected for the model. As a result, both a simplified model (GT and T) and an extended model (GT, T, L, B, EP, DWT) were developed. Ordinary Least Squares (OLS) and robust regression methods were applied to both models. In the GT–T model, passenger capacity was predicted with R² ≈ 0.73, while in the extended model, the explanatory power improved to R² ≈ 0.75. The robust regression approach limited the influence of outliers, but overall results were very similar to those of the OLS model. Diagnostic tests confirmed that the assumptions of the models were met and that the error distributions were close to normal. These findings suggest that both simplified and extended regression models can serve as effective and reliable tools for passenger capacity estimation in engineering applications.