Over the last twenty years, the number of passenger vessels in Adriatic has increased and, due to it, the number of possibilities of maritime accidents has risen as well. In this doctoral dissertation qualitative and quantitative models of risk assessment have been based on analysis of causes of maritime accidents on small passenger vessels. The qualitative models have been made for these types of accidents: sinking, impact, grounding, collision, fire and explosion. The quantitative models have been made for: collision and impact. Influencing factors of maritime accidents and possible combinations of initial factors that could lead vessel to undesirable condition have been analyzed in order to take proactive measures. Qualitative and quantitative models have been developed using Bayesian networks, enabling a visualization of conditional probability between different initial factors and final outcome. Quantification of the influencing factors has been made using statistical data and expert estimation. Quantitative models have been validated by: qualitative feature test, sensitivity analysis and comparison between results and accidental report. By analyzing this and developing the model with the help of Bayesian networks, it is possible to proactively influence the safety of the observed vessels. Based on the results obtained, following the researches, preventive measures have been proposed.