In this project, we intend to develop a decision support system for an autonomous maritime vehicle interacting with other maritime entities. Ships/vessels usually have to sail in and out of cluttered environments while avoiding collision with other vessels. Intent prediction and collision avoiding behavior in manned surface vessels is cognitive and also depends hugely on various visual/audio signals, along with an ability to communicate with neighboring vessels. Predicting intent of other vessels is an even more difficult problem in autonomous vessels. We plan to create a logical model for classifying and identification of maritime entities with the goal to infer the intent of those entities without direct communication or coordination. We propose a hierarchical data-driven approach to predicting intent of maritime entities in collision risk scenarios and otherwise. In this approach, we break down the problem of intent prediction into a series of classification problems. This project is expected to lead to improved decision making models for autonomous systems with limited protocols and without coordination.