Identifying buzz in social media: a hybrid approach using artificial bee colony and k-nearest neighbors for outlier detection

The exponential growth in the use of social media has not only impacted the way individuals communicate and interact but has also opened new avenues for various domains including health care, marketing, e-commerce, e-governance and politics to name a few. It has been further seen that such engagements result in huge amount of user-generated content (UGC) from both individuals and organizations combined. This UGC can be analyzed in multiple ways to mine useful information. One such popular domain that uses this information is content buzz/popularity. The content shared on social media platforms becomes popular and subsequently viral when shared and propagated by a larger audience at a faster pace. Organizations are leveraging this power of social media in the domain of content buzz and virality by employing various buzz monitoring techniques to boost the reach of their content. This study thus proposes a hybrid artificial bee colony approach integrated with k-nearest neighbors to identify and segregate buzz in Twitter. A set of metrics comprising of created discussions, increase in authors, attention level, burstiness level, contribution sparseness, author interaction, author count and average length of discussions are used to model the buzz. The proposed approach considers the buzz discussions as outliers deviating from the normal discussions and identifies the same using the proposed hybrid bio-inspired approach. Findings may be useful in domains like e-commerce, digital and influencer marketing to explore the factors that might create buzz along with the difference between the impact of buzz and normal discussions on the consumers.