Identification of power users in twitter for tweet recommendation

Abstract

Social networks have become the most important platform of opinion spreading. Using topic modeling, it is easy to determine the topic of discussion related to information spread. The rate of information generated in twitter is exponential and tremendous. Therefore a suitable algorithm for topic analysis in real-time is the need of the hour. The results are more accurate when unigrams and bigrams are considered during tweet clustering. From the topic detected tweets, it is comfortable to analyze, recognize patterns, construct models and predict user behaviour. Tweet Analysis can help understand user behaviour and help service providers improve their user experience. This idea is strategically important for companies interested in obtaining feedback for their products, for brand endorsements, merchandising etc. By analyzing the content from social media like twitter, companies can decide their target groups and select influential users from these platforms to promote or endorse their brands. It is also essential to identify how these influential users are influenced. This work proposes unsupervised approaches for modeling power users in twitter. The power users tend to impact the influential users and their social behaviour. This social impact of influence is identified using topic modeling. The tweets are recommended to regular users with respect to the power user they are connected to. Further, this power user model is updated dynamically for effective tweet recommendation and therefore, every user might have different power users for varied social interests. newline