IF-THEN rules in neural networks for classification

Authors:
D. Rutkowska
Affiliations:
Technical University of Czestochowa, Armii Krajowej
Venue:
CIMCA '05 Proceedings of the International Conference on Computational Intelligence for Modelling, Control and Automation and International Conference on Intelligent Agents, Web Technologies and Internet Commerce Vol-2 (CIMCA-IAWTIC'06) - Volume 02
Year:
2005

Citing 0
Cited 2

Use of Circle-Segments as a Data Visualization Technique for Feature Selection in Pattern Classification

Neural Information Processing
Use of the circle segments visualization technique for neural network feature selection and analysis

Neurocomputing

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents a new approach to create neural networks for classification. The classical type of multilayer perceptron (MLP) neural network is considered, but the main idea is to explain the performance of the network with regard to the rule base knowledge representation. The rule base and the classical neural network that works according to these rules is constructed by analysing the data visualisation, and taking into account only the most significant attributes. Creating the rules, new classes are distinguished from the overlapping classes, in order to avoid misclassifications. This approach is illustrated on the well known iris classification problem but it can be applied to much more sophisticated data, and of course not linearly separable.