Title :
Predicting Number of Unsupervised Clusters by Supervised Function
Author :
Srivisal, Chouvanee ; Lursinsap, Chidchanok
Author_Institution :
Dept. of Math., Chulalongkorn Univ., Bangkok, Thailand
Abstract :
The clustering is one of popular technique for separating the similar data into the same group.The problem of this technique is "How to find the real number of group in the data?". So, In this paper we try to find the solution that can guess the number of group by automatic. However, this number is very difficult to specify if the data space is in a very high dimension. Here, the problem of predicting number of clusters is transformed to the problem of constructing a function by using a supervised neural network and locating all zero gradients on this function. The number of locations having zero gradients is equal to the number of clusters. The proposed technique correctly predicts the number of clusters when it is tested with several testing sets.
Keywords :
data mining; learning (artificial intelligence); neural nets; pattern clustering; real number system; similar data separation; supervised neural network; unsupervised cluster number prediction by supervised function; zero gradient location; Bayesian methods; Clustering algorithms; Computational intelligence; Distance measurement; Euclidean distance; Mathematics; Neural networks; Supervised learning; Telecommunication traffic; Testing; Clustering; Predict the number of cluster; Supervised Function;
Conference_Titel :
Computational Sciences and Optimization, 2009. CSO 2009. International Joint Conference on
Conference_Location :
Sanya, Hainan
Print_ISBN :
978-0-7695-3605-7
DOI :
10.1109/CSO.2009.132
