Finding event-specific influencers in dynamic social networks

Reputation models are widely in use today in commercial transaction (ebay), product review (amazon, epinions), and news commentary websites (slashdot). The purpose of these reputation models is to provide behavioral or informational data for future users to determine whether or not he or she will trust the data. These models are dependent on explicit feedback mechanisms where users rate product, other users, or information. However, for many popular social network information sources on the web, no such explicit feedback systems exist where users rate information in order for consumers of this information to be able to judge the trustworthiness of the data source or the data itself. Here I describe the layers of the problem of determining reputation among users or data during events discussed on social networks, and evaluate data and network analysis methods from varying disciplines that may implicitly infer user or data reputation based on metadata, user relationships and user actions in social networks. I demonstrate that the HITS algorithm is not effective at finding influential users, and propose a new algorithm and demonstrate its effectiveness for finding influential users during an event