Analyzing social media communications for correlation with freeway vehicular traffic

Abstract

Participatory sensing of traffic, where Internet users share traffic situations either observed by them or automatically sensed by sensors on their smartphones, has become an established means of collecting real-time traffic information (e.g., Google Traffic and Waze). Recent studies have identified strong correlation between social media activity and traffic patterns in the New York area. Interestingly, social media communication leads road traffic. This project computes similar correlations in southern California with an aim to increase the time-resolution of vehicle traffic prediction as near real-time message volume information is available from online social networks. This study collects the number of public messages sent from southern California on the Twitter social media platform. Messages related to several topics are collected. Vehicle flow data at major freeways in the Los Angeles region as measured by Caltrans' extensive network of inductive loop sensors from Caltrans Performance Measurement System (PeMS). The correlation between the two datasets — social media communications and vehicle traffic — is calculated between these datasets. The strengths of these correlations are used to describe the relationship between social media traffic and vehicular traffic in different regions in the southern California. Results show that social media communication is correlated to traffic sensor data with both showing a periodic pattern with approximately the same period. The relationship between social media and vehicular traffic is similar in different roadways in the same region and also holds across regions. Correlation of traffic with volume of social media communications is maximized with a lag of approximately 3 hours (i.e., social media changes appear before traffic volume changes). If further study can also show that there is predictive value in this observation, the time resolution of traffic prediction can be increased by adapting it to real-time changes in social media message volume.