Fuzzy Distance Measure Based Affinity Propagation Clustering

Affinity Propagation (AP) is an effective algorithm that find exemplars repeatedly exchange real valued messages between pairs of data points. AP uses the similarity between data points to calculate the messages. Hence, the construction of similarity is essential in the AP algorithm. A common choice...

詳細記述

保存先:
書誌詳細
主要な著者: Al-Akash, Omar Mahmoud Nayef, Syed Ahmad, Sharifah Sakinah, Azmi, Mohd Sanusi
フォーマット: 論文
言語:English
出版事項: Research India Publications 2018
主題:
オンライン・アクセス:http://eprints.utem.edu.my/id/eprint/22848/2/non-indeks%20Omar%20Akash%2091_62414-IJAER%20ok%201501-1505.pdf
http://eprints.utem.edu.my/id/eprint/22848/
https://www.ripublication.com/ijaer18/ijaerv13n2_91.pdf
タグ: タグ追加
タグなし, このレコードへの初めてのタグを付けませんか!
その他の書誌記述
要約:Affinity Propagation (AP) is an effective algorithm that find exemplars repeatedly exchange real valued messages between pairs of data points. AP uses the similarity between data points to calculate the messages. Hence, the construction of similarity is essential in the AP algorithm. A common choice for similarity is the negative Euclidean distance. However, due to the simplicity of Euclidean distance, it cannot capture the real structure of data. Furthermore, Euclidean distance is sensitive to noise and outliers such that the performance of the AP might be degraded. Therefore, researchers have intended to utilize different similarity measures to analyse the performance of AP. nonetheless, there is still a room to enhance the performance of AP clustering. A clustering method called fuzzy based Affinity propagation (F-AP) is proposed, which is based on a fuzzy similarity measure. Experiments shows the efficiency of the proposed F-AP, experiments is performed on UCI dataset. Results shows a promising improvement on AP.