The design of new ships is a process that requires knowledge of several aspects of naval architecture and marine engineering. During the early design stage, one of the first issues that designers should face is the preliminary estimation of the vessel's main dimensions, respecting the desiderata of the ship owner. Therefore, it is relevant to provide designers with suitable tools that may help estimate the principal dimensions, consider conventional methods and investigate the applicability of modern techniques based on machine learning. The present work focuses on applying different regression techniques to a database of RoPax vessels, finding mathematical instruments to evaluate the ship's main dimensions. Conventional regression techniques are first investigated here to compare with the existing formulae provided by other databases. The study is then extended by applying multiple linear regression and forest tree algorithms, seeking an improvement of conventional formulations available in the literature. The results highlight how the most modern regression techniques allow for better coverage of the design space, allowing the use of more than one input to obtain the final dimensions.
