Extracting Knowledge from Databases and ANNs with Genetic Programming: Iris Flower Classification Problem

In this chapter, we present an application of Genetic Programming (GP) in the field of data mining and extraction of Artificial Neural Networks (ANN) rules. To do this, we will use its syntactic properties to obtain high level expressions that represent knowledge. These expressions will have different types as there is the need at each moment: we will obtain different expressions like IF-THEN-ELSE rules, mathematical relations between variables or boolean expressions. In this chapter, we will not only apply GP to solve the problem, but we will try different modifications and This chapter appears in the book, Intelligent Agents for Data Mining and Information Retrieval, edited by Masoud Mohammadian. Copyright © 2004, Idea Group Inc. Copying or distributing in print or electronic for s without written permission of Idea Group Inc. is prohibited. 701 E. Chocolate Avenue, Suite 200, Hershey PA 17033-1240, USA Tel: 717/533-8845; Fax 717/533-8661; URL-http://www.idea-group.com IDEA GROUP PUBLISHING Iris Flower Classification Problem 137 Copyright © 2004, Idea Group Inc. Copying or distributing in print or electronic forms without written permission of Idea Group Inc. is prohibited. different ways to apply it to solve the problem. We will show how making a data pre-processing we can obtain better results than using the original values. That is, by adding a little knowledge from the problem we can improve the performance of GP. INTRODUCTION In the world of Artificial Intelligence (AI), the extraction of knowledge has been a very useful tool for many different purposes, and it has been tried with many different techniques. Here, we will show how we can use Genetic Programming (GP) to solve a classification problem from a database, and we will show how we can adapt this tool in two different ways: to improve its performance and to make it possible to detect errors. Results show that the technique developed in this chapter opens a new area for research in the field, extracting knowledge from more complicated structures such as Artificial Neural Networks (ANNs). BACKGROUND Genetic Programming and Artificial Neural Networks Genetic Programming (GP) (Koza, 1992) is an evolutionary method used to create computer programs that represent approximate or exact solutions to a problem. This technique allows for the finding of programs with the shape of a tree, and, in its most common application, those programs will be mathematical expressions combining mathematical operators, input variables, constants, decision rules, relational operators, etc. All of these possible operators must be specified before starting the search. So, with them, GP must be able to build trees with the objective of finding the desired expression which models the relation between the input variables and the desired output. This set of operators is divided into two groups: the terminal set contains the operators which cannot accept parameters, like variables or constants; and the function set, which contains the operators, such as add or subtract, which need parameters. Once the terminal and non-terminal operators are specified, it is possible to establish types. Each node will have a type, and the construction of child expressions needs to follow the rules of the nodal type (Montana, 1995). GP creates automatic program generation by means of a process based on the evolution theory of (Darwin, 1864), in which, after subsequent generations, 15 more pages are available in the full version of this document, which may be purchased using the "Add to Cart" button on the product's webpage: www.igi-global.com/chapter/extracting-knowledgedatabases-anns-genetic/24160?camid=4v1 This title is available in InfoSci-Books, InfoSci-Database Technologies, Library Science, Information Studies, and Education, InfoSci-Library and Information Science, InfoSciComputer Science and Information Technology, Science, Engineering, and Information Technology, InfoSci-Select, InfoSci-Select. Recommend this product to your librarian: www.igi-global.com/e-resources/libraryrecommendation/?id=1