Find the butterfly: a social media based arterial incidents detection and causality analysis system

Kaiqun Fu, Weisheng Zhong, Chang-Tien Lu, Arnold Boediardjo

Abstract

Traditional statistical analysis on speed, volume, and occupancy has dominated the field of Arterial Incident Management Study (AIMS). However, few previous works have focused on investigating into the causality of the incidents. In this paper, we present ButterFly, a social media based arterial incident detection and analysis system. The proposed system is dedicated to identify the traffic incident from a novel perspective and discover causalities between traffic incidents. The main functionalities of the proposed system include: 1) Traffic incident detection based on userinput social media contents, 2) Transportation incidents storyline generation, and 3) Traffic incidents causalities analysis and visualization. We demonstrate the system by considering the Washington DC area as our experimental environment. ButterFly is targeted to provide effective and convenient real-time and historical traffic incidents analysis interfaces for transportation management agencies and academies. Our proposed system, integrated with multiple social media resources, can greatly broaden the visions for traffic incidents analysis.