The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #9760

Title:Generating Fuzzy Classi
Date:Sep 1997
Abstract:A method of extracting intuitive knowledge from neural network classifiers is presented in this thesis. Rule induction is an important key to the solution of the knowledge acquisition bottleneck in the development of knowledge-based expert systems. The particular algorithm obtains crisp rules in the form of logical implications which approximately describe the neural network mapping. the number of produced rulescan be selected using an uncertainty margine parameter. Three classification problems, the IRIS database, the ex-or function and the weather forecast have been tested. The result of the experiments in this thesis shows that noisy data cause hight classification error, thus the neural network is a necessary element because it provides filtration of the training data.

[Search These Pages] [DAI Home Page] [Comment]