Using text mining and sentiment analysis for online forum hotspot detection

less than 1 minute read

by Nan Li and Desheng Dash Wu (Decision Support Systems)

The authors combine the following five features to detect forum hotspots using either (i) K-means clustering or (ii) support vector machines:

  1. number of total topic posts (in every forum)
  2. the average number of responses to these posts
  3. the average forum sentiment (using a _simple_ bag of words approach)
  4. the percentage of positive postings
  5. the percentage of negative postings (which is in fact dependent on 4 ;)
They identify a cluster center for every forum and identify the sub-forums closest to this theoretical center as hotspots.