Combination of singular value decomposition and K-means clustering methods for topic detection on Twitter
Authors:
- 1 October 2015
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 123-128
Abstract
Online social media are growing very rapidly in recent years, such as Twitter. Even the interaction and communication in the social media can reflect on the events of the real world. This causes the value of the information increasing significantly. However, the huge amount of the information requires a method of automatically detecting topics, one of which is the K-means Clustering. Moreover, the large dimensions of data become obstacles. So, we used singular value decomposition (SVD) to reduce the dimension of the data prior to the learning process using the K-means Clustering. The accuracy of the combination of SVD and K-means Clustering methods showed comparative results, while the computation time required is likely to be faster than the method of K-means Clustering without any reduction in advance.Keywords
This publication has 7 references indexed in Scilit:
- Sensing Trending Topics in TwitterIEEE Transactions on Multimedia, 2013
- Probabilistic topic modelsCommunications of the ACM, 2012
- Matrix Computations2012
- Latent semantic analysisAnnual Review of Information Science and Technology, 2004
- Topic Detection and Tracking2002
- Learning the parts of objects by non-negative matrix factorizationNature, 1999
- Indexing by latent semantic analysisJournal of the American Society for Information Science, 1990