Identifying Traffic Event Types from Twitter by Multi-Label Classification

Abstract

Social networks have been recently employed as a source of information for traffic event detection, with particular reference to traffic incidents. However, extracting meaningful and actionable knowledge from user-generated content is a complex endeavor, because social media texts are usually short, informal, with a lot of abbreviations, jargon, slang, idioms, and usually belong multiple labels or topics, so the classification of an only topic is difficult to be predicted. In this paper, we present an approach for classifying traffic event from the social network Twitter. The system fetches publications from Twitter (i.e., tweets) according to several search criteria; processes tweets, by applying text mining techniques; and finally performs the multi-label classification of tweets. The aim is to assign the appropriate class labels to each tweet, as related to traffic event types. The proposed model was employed in the Brazilian city of Porto Alegre, allowing for classification of traffic event types. Extensive experiments show the effectiveness of our proposed method for multi-label classification of Twitter posts, achieving high precision at substantial overall coverage.