Comparative performances of stochastic competitive evolutionary neural tree (SCENT) with neural classifiers
Author
Pensuwon, W.
Adams, R.G.
Davey, N.
Attention
2299/830
Abstract
A stochastic competitive evolutionary neural tree (SCENT) is described and evaluated against the best neural classifiers with equivalent functionality, using a collection of data sets chosen to provide a variety of clustering scenarios. SCENT is firstly shown to produce flat classifications at least as well as the other two neural classifiers used. Moreover its variability in performance over the data sets is shown to be small. In addition SCENT also produces a tree that can show any hierarchical structure contained in the data. For two real world data sets the tree captures hierarchical features of the data.