TrafficCast International's fusion process includes model integration, information filtering, and data normalization. As stated before, TrafficCast International uses a combination of statistical model, heuristic model, simulation model, and Dynamic Traffic Assignment model. To use which model or a combination of models is dependent on the application needs and the availability of data. In some cases, the output of statistical model and/or heuristic model is used as the initial input to simulation model and/or DTA model to generate dynamic traffic information and forecasting to support a variety of applications.

TrafficCast International uses a rigorous Quality Assurance and Quality Control (QA/QC) mechanism to calibrate and verify the model output to ensure the prediction accuracy. For instance, by using fuzzy logic TrafficCast International develops an online calibration procedure to compare the predictive speed generated by model to the speed collected by DOT speed sensors in real-time to verify its prediction accuracy. If this discrepancy is over a pre-set threshold, the model “learns” the mistake and adjusts itself internally to improve its prediction. This process is implemented repeatedly very 5 minutes for markets where real-time speed is available. For markets without speed sensors but have traffic cameras, a web-based operator interface allows operators to provide input based on what he/she sees on the camera for model to change its prediction behavior.