The contemporary practice of estimations on allision frequencies between watercrafts and offshore installations is associated with a degree of conservatism due to modeling simplicity. This implies that the allision frequency may be estimated higher than the actual value, now and in the future. This dissertation proposes an alternative method of estimation of allision frequencies building on adaptive networks, Bayesian Belief Networks. A showcased model in terms of the allision frequency of the Halfdan B oil and gas offshore installation complex is elaborated and quantified. The vast amounts of AIS data can be applied for detailed modeling of maritime traffic patterns and new patterns forming may be predicted well in advance to anticipate the decrease or increase of allision frequencies of existing offshore installation infrastructure. This dissertation concludes that when more advanced modeling is applied to allision frequency estimation the frequency is in fact different as opposed to contemporary practices.