The safety of navigation is of global importance. Good organisation, surveillance and management of maritime traffic contribute significantly to the safety of any maritime region. Although great efforts have already been made to improve safety, maritime accidents still occur. By analysing traffic and ship data, it is possible to identify areas where there is a higher risk due to increased traffic density. This makes it possible to measure the efficiency and safety of maritime traffic in a monitored region, considering the ship collision probability. Maritime traffic density is the number of ships passing through the monitored area in a given time period. The challenge for coastal maritime regions is the variation in traffic density and mode characteristics. Another challenge is how this variation in shipping routes affects the ship collisions probability or other accidents at sea. Traffic density can be measured using data from Automatic Identification System – AIS. The information provided by AIS is real-time data designed to improve maritime safety and security. AIS data can also be used for scientific research purposes to improve maritime safety by developing predictive models for collisions in a research area. The AIS data are received and stored in "raw format" - National Marine Electronic Association – NMEA files. These files must be processed, i.e., decoded. Only decoded NMEA files can be used in the model and to determine collision probability. The objective of the model in this research is to analyse the traffic density and determine the collision probability based on the available AIS data for the research area. The proposed model is applied to the Split maritime area. However, it can be applied to any navigational area for which the AIS data set exists. The proposed architecture of the collision probability model includes the following 5 steps: 1. AIS data collecting, processing and verification; 2. Decoding and cleaning-up of AIS data; 3. Route data set creation; 4. Monte Carlo simulations – MC; 5. Learning, validating and testing the proposed Bidirectional Long Short – Term Memory Artificial Neural Network – Bi-LSTM.