Forecasting Word Model: Twitter-based Influenza Surveillance and Prediction

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 76–86, Osaka, Japan, December 11-17 2016. Forecasting Word Model: Twitter-based Influenza Surveillance and Prediction Hayate ISO, Shoko WAKAMIYA, Eiji ARAMAKI Nara Institute of Science and Technology {iso.hayate.id3,wakamiya,aramaki}@is.naist.jp Abstract Because of the increasing popularity of social media, much information has been shared on the internet, enabling social media users to understand various real world events. Particularly, social media-based infectious disease surveillance has attracted increasing attention. In this work, we specifically examine influenza: a common topic of communication on social media. The funda- mental theory of this work is that several words, such as symptom words (fever, headache, etc.), appear in advance of flu epidemic occurrence. Consequently, past word occurrence can contribute to estimation of the number of current patients. To employ such forecasting words, one can first estimate the optimal time lag for each word based on their cross correlation. Then one can build a linear model consisting of word frequencies at different time points for nowcasting and for forecasting influenza epidemics. Experimentally obtained results (using 7.7 million tweets of August 2012 – January 2016), the proposed model achieved the best nowcasting performance to date (correlation ratio 0.93) and practically sufficient forecasting performance (correlation ratio 0.91 in 1-week future prediction, and correlation ratio 0.77 in 3-weeks future prediction). This report reveals the effectiveness of the word time shift to predict of future epidemics using Twitter. 1 Introduction The increased use of social media platforms has led to wide sharing of personal information. Espe- cially Twitter, a micro-blogging platform that enables users to communicate by updating their status using 140 or fewer characters, has attracted great attention of researchers and service developers because Twitter can be a valuable personal information resource. The feasibility of such approaches, known as social sensors, has been demonstrated in various event detection systems such as earthquakes (Sakaki et al., 2010), outbreaks of disease (Chew and Eysenbach, 2010), and stock market fluctuations (Bollen et al., 2011). Among the applications mentioned above, this study particularly examines detection of seasonal influenza epidemics because the influenza detection is a popular application of Twitter. To date, more than 30 Twitter-based influenza detection and prediction systems have been developed worldwide (Charles-Smith et al., 2015). Although the detailed functions of these systems differ, they share the underlying assumption that the flu spreading in the real world is immediately reflected to the tweets. Therefore, most systems have simply aggregated counts of daily flu-related tweets to obtain the current patient status (Aramaki et al., 2011; Collier et al., 2011; Chew and Eysenbach, 2010; Lampos and Cristianini, 2010; Culotta, 2013; Paul et al., 2014). Their typical materials are presented as shown below. • I got a flu I can not go to school for the rest of the week • I was diagnosed with a high fever. Maybe flu :( Although the former tweet is described by an actual influenza patient, the latter one merely expresses a suspicion of flu. From a practical …

Content truncated to 500 words