Predicting Short-Term Traffic Flow Congestion on Urban Motorway Networks

The project developed a short-term traffic flow prediction model using machine learning techniques. The team collected traffic data from Canada and Europe and trained the model to predict traffic flow up to 15 minutes ahead. Various techniques, including neural networks and decision trees, were used to improve the accuracy of the predictions.

The problem

The Smart-Traffic system for real-time traffic prediction has in place a method for predicting congested road-vehicle traffic on a given roadway within a region.

In particular, the computer-implemented method utilizes real-time traffic images from traffic cameras for the input of data and utilizes computer processing and machine learning to model a predictive level of congestion within a category of low congestion, medium congestion, or high congestion. By implementing machine learning in the comparison of exemplary images and administrator review, the computer processing system and method steps can predict a more efficient real-time congestion prediction over time.

Source

The project outcomes

There are two phases to the project outcomes:

• Obtaining API traffic image data for a new city in Europe or North America or setting up an end-to-end solution.
• Enhancing the model’s predictive accuracy by removing noise such as trees, dual lanes, etc. that may affect the camera’s object focus.

The podcast related to the traffic problem