Modeling the Spread of Information on Twitter

Compartmental models have been used in epidemiology for many years to study the spread of infectious diseases throughout the world. In this thesis, we are recreating and extending the work done by others in [1] to apply one of these models, the SEIZ model, to the spread of news and rumors on Twitter. After deriving the model and discussing its background, we obtained data regarding 6 events, 3 real news stories and 3 rumors. We showed that the method used to minimize the error between the model and the actual data was quite accurate, and that the model was able to work very early on in a story or with limited information. We also attempted to find several combinations of parameters which could distinguish the stories between news and rumors, but no consistent results were found. Finally, we restricted the amount of data fed into the model, and took a look at its ability to estimate the number of tweets in the future.