Knowledge Discovery Using Genetic Programming with Rough Set Evaluation

An important area of KDD research involves development of techniques which transform raw data into forms more useful for prediction or explanation. We present an approach to automating the search for "indicator functions" which mediate such transformations. The fitness of a function is measured as its contribution to discerning different classes of data. Genetic programming techniques are applied to the search for and improvement of the programs which make up these functions. Rough set theory is used to evaluate the fitness of functions. Rough set theory provides a unique evaluator in that it allows the fitness of each function to depend on the combined performance of a population of functions. This is desirable in applications which need a population of programs that perform well in concert and contrasts with traditional genetic programming applications which have as there goal to find a single program which performs well. This approach has been applied to a small database of iris flowers with the goal of learning to predict the species of the flower given the values of four iris attributes and to a larger breast cancer database with the goal of predicting whether remission will occur within a five year period. Introduction An important area of KDD research involves development of techniques which transform raw data into forms more useful for prediction or explanation. We present an approach to automating the search for "indicator functions" which mediate such transformations. The fitness of a function is measured as its contribution to discerning different classes of data. Genetic programming techniques are applied to searching for and improving the programs which make up these functions. Rough set theory is used to evaluate the fitness of functions. Rough set theory provides a unique evaluator in that it allows the fitness of each function to depend on the combined performance of a population of functions. This is desirable in applications which need a population of programs that perform well in concert and contrasts with traditional genetic programming applications which have as there goal to find a single program which performs well. This approach has been applied to a small database of iris flowers with the goal of learning to predict the species of the flower given the values of four iris attributes (Fisher’s iris data reproduced in Salzberg, 1990) and to a larger breast cancer database (breast cancer data reproduced in Salzberg, 1990) with the goal of predicting whether remission will occur within a five year period. The process begins by applying a population of randomly-generated programs to elements of a database. Program results are placed in a matrix and evaluated to obtain a measure of each program’s fitness. These fitness values are used to determine which programs will be kept and used in breeding the next generation. This process continues for a specified number of generations and is illustrated in Figure 1. Page 254 Knowledge Discovery in Databases Workshop I993 AAAI-93 From: AAAI Technical Report WS-93-02. Compilation copyright © 1993, AAAI (www.aaai.org). All rights reserved. I Population of programs P = {pl,p2,...,pn} Reproduction Crossover Mutation Execution Knowledge Base Application Data: Program primitives Attribute value matrix P x Data Significance of programs Minimal program sets I =@ I Figure 1: Genetic Programming Rank ordered programs