Department of Computer Engineering
S E M I N A R
Modelling of Social Cascades with Recurrent Neural Networks
Computer Engineering Department
Social networking platforms on the Internet are considered to be ideal for viral marketing strategies and can offer a competitive edge to companies in the marketing world. Viral marketing utilizes social cascade, which is alluring to marketers as it is a relatively costless promotion method and relies only on users sharing information voluntarily, to spread its influence among users. The aim is to find the most influential users to promote the advertisement and ensure that its spread across the network reaches as many people as possible. The main task is to keep the number of such seed users minimal while obtaining the maximum distribution rate in the network. In order to extract influential users, we model the social network onto a social graph where each user is a node and interaction between two users establishes an edge that connects the two. Then, the Google PageRank algorithm is applied which ranks the users based on their interactions with others, i.e. their influence over the network. Finally, we use recurrent neural networks such as Hopfield networks to provide us with a simulation environment to observe the word of mouth process in the social network.
DATE: 16 April, 2012, Monday @ 15:40