Modeling spread of word of mouth on Twitter

Abstract

Twitter is a popular word-of-mouth microblogging and online social networking service. Our study investigates the diffusion pattern of the number of mentions, or the number of times a topic is mentioned on Twitter, in order to provide a better understanding of its social impacts, including how it may be used in marketing and public relations. After an extensive literature review on diffusion models and theories, we chose the Bass diffusion model, because it allows us to achieve a relatively good estimation for the diffusion pattern of a trending topic. Furthermore, we extend the Bass model in two ways: (1) incorporating the number of mentions from influential users on Twitter; (2) aggregating the hourly data observations into daily data observations. Both extensions significantly improve the model's ability to predict the total number of mentions and the time of highest mentions. In the future, we hope to extend the applications of our study by incorporating external data from the news and other sources, to provide more comprehensive information about what people are saying and thinking. We also hope to analyze the data in terms of demographics and user networks, to potentially predict everything from new product introduction to conversations about defective products.