The use of data-driven models has become increasingly important in recent years. Efficient prediction of ship performance (e.g., power requirements) and ship responses is required to monitor ship conditions, assess ship safety and comply with regulations. This thesis deals with the analysis and assessment of data-based models and compares them with physics-based models.  Different deep learning (DL) models and the associated algorithms for predicting the power requirements, fuel consumption and motions of ships under different operating conditions were analyzed. Model test data, full-scale measurements on board ships and numerical simulation results were used to validate the data-based models.  The results of this work show that almost all used data-based models can predict the engine power and ship motions with good accuracy. The data-based model ‘AutoML’ (Automated Machine Learning) was able to capture ship motions with the higher accuracy. The present work is a semi-cumulative dissertation comprising five scientific papers. The articles were published in peer-reviewed journals and conference proceedings.