Intuitionistic fuzzy C-regression by using least squares support vector regression

•We developed a novel clustering algorithm for solving actual clustering problems.•The novel clustering algorithm improves conventional fuzzy c-regression model.•Empirical results indicate that the proposed clustering algorithm has superior performance. This paper proposes a novel intuitionistic fuz...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Expert systems with applications Ročník 64; s. 296 - 304
Hlavní autoři: Lin, Kuo-Ping, Chang, Hao-Feng, Chen, Tung-Lian, Lu, Yu-Ming, Wang, Ching-Hsin
Médium: Journal Article
Jazyk:angličtina
Vydáno: Elsevier Ltd 01.12.2016
Témata:
Fuzzy c-regression model
Intuitionistic fuzzy sets
Least squares support vector regression
Sammon mapping
Intuitionistic fuzzy sets
Sammon mapping
Least squares support vector regression
Fuzzy c-regression model
ISSN:0957-4174, 1873-6793
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:•We developed a novel clustering algorithm for solving actual clustering problems.•The novel clustering algorithm improves conventional fuzzy c-regression model.•Empirical results indicate that the proposed clustering algorithm has superior performance. This paper proposes a novel intuitionistic fuzzy c-least squares support vector regression (IFC-LSSVR) with a Sammon mapping clustering algorithm. Sammon mapping effectively reduces the complexity of raw data, while intuitionistic fuzzy sets (IFSs) can effectively tune the membership of data points, and LSSVR improves the conventional fuzzy c-regression model. The proposed clustering algorithm combines the advantages of IFSs, LSSVR and Sammon mapping for solving actual clustering problems. Moreover, IFC-LSSVR with Sammon mapping adopts particle swarm optimization to obtain optimal parameters. Experiments conducted on a web-based adaptive learning environment and a dataset of wheat varieties demonstrate that the proposed algorithm is more efficient than conventional algorithms, such as the k-means (KM) and fuzzy c-means (FCM) clustering algorithms, in standard measurement indexes. This study thus demonstrates that the proposed model is a credible fuzzy clustering algorithm. The novel method contributes not only to the theoretical aspects of fuzzy clustering, but is also widely applicable in data mining, image systems, rule-based expert systems and prediction problems.
ISSN:0957-4174
1873-6793
DOI:10.1016/j.eswa.2016.07.040