Analyzing the Accuracy of Historical Average for Urban Traffic Forecasting using Google Maps

Urban traffic management uses increasingly sophisticated methods to overcome the many challenges involved in the development of traffic forecasting solutions. The main challenge is the acquisition of real-time large-scale urban traffic data at a sufficient spatio-temporal resolution. This is a challenge mainly because of the high financial cost that the installation of a large number of sensors would incur. This paper addresses this challenge by leveraging 'real-time' Google Traffic maps which show the state of the traffic on different road segments using four different colors. Since Google Traffic maps are provided in the form of rendered images, we apply image processing on Google Traffic maps to extract traffic data that are suitable for processing and analysis. The traffic data obtained from Google Traffic are validated with a traffic data set collected by sensors installed in Paris. Then, using data gathered through our Google Traffic-based method for several roads in Rabat (Morocco), we evaluate the accuracy of traffic prediction based on historical average. The overall accuracy reaches 74.9% on week days and 83.3% on weekend days. Further, a more detailed study by type of road and by time period was conducted, showing an overall accuracy of 95.8% in fluid traffic situations.