Abstract
In this paper, a generalized competitive agglomeration (CA) clustering algorithm called entropy index constraints competitive agglomeration (EICCA) is proposed to avoid the drawback that the fuzziness index m in the CA must be fixed to be 2. The proposed EICCA is inspired by a basic fuzzy clustering algorithm called entropy index constraints fuzzy C-means (EIC-FCM), which is comparable to fuzzy C-means (FCM) in clustering performance but completely different from the FCM in the use of entropy index constraints with very clear physical meaning instead of the original constraints in the FCM. With the help of the EIC-FCM, the generalized competitive agglomeration algorithm EICCA is developed by introducing a competition term into the EIC-FCM’s objective function, which is similar to the CA by introducing a competition term into the FCM’s objective function. Our theoretical analysis and empirical results indicate that the EICCA can effectively find the optimal number of clusters for a dataset to be clustered, with more flexible index choices than the CA having the fuzziness index m = 2 only.
Original language | English |
---|---|
Pages (from-to) | 1945-1969 |
Number of pages | 25 |
Journal | International Journal of Machine Learning and Cybernetics |
Volume | 8 |
Issue number | 6 |
DOIs | |
Publication status | Published - 1 Dec 2017 |
Keywords
- Competitive agglomeration
- Entropy index
- Fuzziness index
- Fuzzy clustering
- Hierarchical clustering
- Partitional clustering
ASJC Scopus subject areas
- Software
- Computer Vision and Pattern Recognition
- Artificial Intelligence