Link-Anomaly Detection in Twitter Streams

Rapid growth of social network gives emergence to the detection of emerging topics. The information exchanged over social network post not only includes text but also images, URLs and videos therefore conventional frequency based appropriate in this context. By taking into consideration the links between users that are generated dynamically through replies, mentions, and retweets are included. This paper highlights the analysis of a probability model that mention the behavior of a social network user. This model is used to detect the anomalies emerged. From hundreds of users anomaly scores are aggregated. In the proposed system it is only based on replay/mention relationship and is experiment zed with in real datasets gathered from twitter