In a connected travel environment, the transportation sector has become a major producer of “Big Data” gathered through a variety of sensing capabilities from unlimited centralized and decentralized sources (example: cell phones, traffic detectors, CCTV cameras, connected vehicles …etc.). With such data, transportation decision makers aim to develop proactive traffic control and safety monitoring tools. TRANSMART has deployed a series of Artificial Intelligence (AI) techniques to translate such data into analysis and prediction models of congestion and collision formation. In particular, traffic detector and collision data from the Virginia Department of Transportation (VDOT), the Virginia Department of Motor Vehicles (VA DMV), and the Korea Expressway Corporation (KEC) have been embedded in a neural network algorithm to predict the occurrences of 80% of collisions in quasi real time. With proper training and cross-validation, speed distributions have been identified as a key measure in order to specify abnormal patterns (i.e. incidents) and extract related characteristics such as type of disruption (i.e. partial versus complete closure) and disruption duration.