This paper involves two approaches for finding the trending topics in social networks that is key-based approach and link-based approach. In conventional key-based approach for topics detection have mainly focus on frequencies of (textual) words. We propose a link-based approach which focuses on posts reflected in the mentioning behavior of hundreds users. The anomaly detection in the twitter data set is carried out by retrieving the trend topics from the twitter in a sequential manner by using some API and corresponding user for training, then computed anomaly score is aggregated from different users. Further the aggregated anomaly score will be feed into change-point analysis or burst detection at the pinpoint, in order to detect the emerging topics. We have used the real time twitter account, so results are vary according to the tweet trends made. The experiment shows that proposed link-based approach performs even better than the keyword-based approach.
展开▼
