4 Conclusion
To overcome the weakness of short text such as its sparsity and a single clustering
algorithm may not be adapted to any type of data structure, this paper proposed a
clustering ensemble algorithm based on CNN for short text. The short text vectors
generated by word2vec are further extracted by CNN to obtain low-dimensional feature
vectors. Then combining with Gini coefficient an index to evaluate the reliability of
clustering is proposed which is called ROC. The CA matrix is redefined on the basis of
considering the reliability of clustering and the diversity within clustering. The ROCCA matrix proposed in this paper is used as the input of the final hierarchical clustering
function, and the final consistency result is obtained. The above contents are the
research results so far. The experiment in this paper is in progress, so no specific data
charts are shown for the time being. In the future work, detailed experiments will be
carried out to obtain sufficient data to support the method in this paper.
References
1. Lin D (1998) An information-theoretic definition of similarity. In: ICML, pp 296–304
2. Ramage D, Heymann P, Manning CD et al (2009) Clustering the tagged web. In:
Second ACM international conference on web search and data mining, ACM, pp 54–63
3. Suo H et al (2006) An improved K-means algorithm for document clustering. J Shandong
Univ (Natural Science) 43(1):60–64
4. Yang Z et al (2014) Online comment clustering based on an improved semantic distance.
J Softw 25(12):2777–2789
5. Zhang Q et al (2016) Short text clustering algorithm combined with context semantic
information. Comput Sci 43(s2):443–446
6. Wang T (2011) CA-tree: a hierarchical structure for efficient and scalable coassociationbased cluster ensembles. IEEE TRANS Syst Man Cybern B Cybern 41(3):686–698
7. Yu Z, Li L, Liu J (2015) Adaptive noise immune cluster ensemble using affinity propagation.
IEEE Trans Knowl Data Eng 27(12):3176–3189
8. Fred ALN, Jain AK (2005) Combing multiple clusterings using evidence accumulation.
IEEE Trans Pattern Anal March Intell 27(6):835–850
9. Huang D, Lai J-H (2015) Combining multiple clusterings via crows agreement estimation
and multi-granularity link analysis. Neurocomputing 170:240–250
10. Strehl A, Gosh J (2005) Cluster ensembles–A knowledge reuse framework for combining
multiple partitions. J Mach Learn Res 3:583–850
11. Li T, Ding C (2008) Weighted consensus clustering. In: Proceedings of the 2008 SIAM
international conference on data mining (SDM), Atlanta, GA, USA, 2008, pp 798–809
12. Yu Z et al (2014) Hybrid clustering solution selection strategy. Pattern Recognit 47
(10):3362–3375
13. Blei DM, Jordan MI (2003) Latent dirichlet allocation. J Mach Learn Res Achieve 993–1022
628
H. Wan et al.
To overcome the weakness of short text such as its sparsity and a single clustering
algorithm may not be adapted to any type of data structure, this paper proposed a
clustering ensemble algorithm based on CNN for short text. The short text vectors
generated by word2vec are further extracted by CNN to obtain low-dimensional feature
vectors. Then combining with Gini coefficient an index to evaluate the reliability of
clustering is proposed which is called ROC. The CA matrix is redefined on the basis of
considering the reliability of clustering and the diversity within clustering. The ROCCA matrix proposed in this paper is used as the input of the final hierarchical clustering
function, and the final consistency result is obtained. The above contents are the
research results so far. The experiment in this paper is in progress, so no specific data
charts are shown for the time being. In the future work, detailed experiments will be
carried out to obtain sufficient data to support the method in this paper.
References
1. Lin D (1998) An information-theoretic definition of similarity. In: ICML, pp 296–304
2. Ramage D, Heymann P, Manning CD et al (2009) Clustering the tagged web. In:
Second ACM international conference on web search and data mining, ACM, pp 54–63
3. Suo H et al (2006) An improved K-means algorithm for document clustering. J Shandong
Univ (Natural Science) 43(1):60–64
4. Yang Z et al (2014) Online comment clustering based on an improved semantic distance.
J Softw 25(12):2777–2789
5. Zhang Q et al (2016) Short text clustering algorithm combined with context semantic
information. Comput Sci 43(s2):443–446
6. Wang T (2011) CA-tree: a hierarchical structure for efficient and scalable coassociationbased cluster ensembles. IEEE TRANS Syst Man Cybern B Cybern 41(3):686–698
7. Yu Z, Li L, Liu J (2015) Adaptive noise immune cluster ensemble using affinity propagation.
IEEE Trans Knowl Data Eng 27(12):3176–3189
8. Fred ALN, Jain AK (2005) Combing multiple clusterings using evidence accumulation.
IEEE Trans Pattern Anal March Intell 27(6):835–850
9. Huang D, Lai J-H (2015) Combining multiple clusterings via crows agreement estimation
and multi-granularity link analysis. Neurocomputing 170:240–250
10. Strehl A, Gosh J (2005) Cluster ensembles–A knowledge reuse framework for combining
multiple partitions. J Mach Learn Res 3:583–850
11. Li T, Ding C (2008) Weighted consensus clustering. In: Proceedings of the 2008 SIAM
international conference on data mining (SDM), Atlanta, GA, USA, 2008, pp 798–809
12. Yu Z et al (2014) Hybrid clustering solution selection strategy. Pattern Recognit 47
(10):3362–3375
13. Blei DM, Jordan MI (2003) Latent dirichlet allocation. J Mach Learn Res Achieve 993–1022
628
H. Wan et al.