Using Twitter data for transit performance assessment: a framework for evaluating transit riders’ opinions about quality of service
operations - performance, planning - service quality, planning - service improvement, planning - travel demand management, technology - passenger information, ridership - perceptions
Topic modeling, Latent Dirichlet allocation (LDA), Sentiment analysis, Transit service performance, Quality of transit service
Social media platforms such as Facebook, Instagram, and Twitter have drastically altered the way information is generated and disseminated. These platforms allow their users to report events and express their opinions toward these events. The profusion of data generated through social media has proved to have the potential for improving the efficiency of existing traffic management systems and transportation analytics. This study complements existing literature by proposing a framework to evaluate transit riders’ opinion about quality of transit service using Twitter data. Although previous studies used keyword search to extract transit-related tweets, the extracted tweets can still be noisy and might not be relevant to transit quality of service at all. In this study, we leverage topic modeling, an unsupervised machine learning technique, to sift tweets that are relevant to the actual user experience of the transit system. Sentiment analysis is further performed based on the tweet-per-topic index we developed, to gauge transit riders’ feedback and explore the underlying reasons causing their dissatisfaction on the service. This framework can be potentially quite useful to transit agencies for user-oriented analysis and to assist with investment decision making.
Permission to publish the abstract has been given by SpringerLink, copyright remains with them.
Haghighi, N.N., Liu, X.C., Wei, R., Li, W., & Shao, H. (2018). Using Twitter data for transit performance assessment: a framework for evaluating transit riders’ opinions about quality of service. Public Transport, Vol. 10, pp. 363-377.