Classification of Traffic Events in Mexico City Using Machine Learning and Volunteered Geographic Information

Abstract

Volunteer geographic information and user-generated content represents a source of updated information about what people perceive from their environment. Its analysis generates the opportunity to develop processes to study and solve social problems that affect the people's lives, merging technology and real data. One of the problems in urban areas is the traffic. Every day at big cities people lose time, money, and life quality when they get stuck in traffic jams; another urban problem derived from traffic is air pollution. In the present approach, a traffic event classification methodology is implemented to analyze VGI and internet information related to traffic events with a view to identify the main traffic problems in a city and to visualize the congested roads. The methodology uses different computing tools and algorithms to achieve the goal. To obtain the data, a social media and RSS channels are consulted. The extracted data texts are classified into seven possible traffic events, and geolocalized. In the classification, a machine learning algorithm is applied.