Traffic congestion in freeways poses a major threat to the economic prosperity of the nation. It not only causes loss of productivity of the workforce but also frustration among the drivers. Traffic incidents such as vehicle crashes, overturned trucks, and stalled vehicles contribute to a significant amount of non-recurrent congestion. Automatic Incident Detection (AID) algorithms have been developed for detecting such incidents in real-time and alerting the drivers. Such incident detection algorithms often rely on generating thresholds based on historical traffic patterns. However, large-scale historical traffic data often poses to be a major challenge in processing these data and automatically generate thresholds for incident detection. This project leverages lambda architecture and cloud computing service to develop an automatic threshold computation module. It is based on the Inter-Quartile Distance (IQD) method for threshold computation and runs on a weekly basis using previous 8 weeks of historical data, amounting to more than 250 GB of traffic data. The incidents detected using the real-time traffic data and the automatically generated thresholds can be used by Traffic Management Centers for real-time incident detection and further analysis such as incident validation and performance tuning.