Effective barge scheduling in the logistic domain requires advanced information on the availability of the port terminals and the maritime traffic in their vicinity. To enable a long-term prediction of vessel arrival times, we investigate how to use the publicly available automatic identification system (AIS) data to identify maritime patterns and transform them into a directed graph that can be used to estimate the potential trajectories and destination points. To tackle this problem, we use a genetic algorithm (GA) to cluster vessel position data. Then, we show how to enhance the process to allow fast computation of incremental data coming from the sensors, including the importance of adding a quad tree structure for data preprocessing. Focusing on a real case implementation, characterized by partially incomplete and noisy AIS data, we show how the algorithm can handle routes intersecting the regions with missing data and the repercussions this has on the route graph. Finally, postprocessing is explained that handles graph pruning and filtering. We validate the results produced by the GA by comparing resulting patterns with known inland water routes for two Dutch provinces followed by the simulation using synthetic data to highlight the strengths and weaknesses of this approach.