Expert Systems in Preprocessing: A Preliminary Study of Supervised Learning with Noise Using C4.5
Abstract
The research work in the System Concepts Section is the study of Pattern Theory. The goal of this work is to have a robust method for finding patterns given a set of data samples and from that underlying pattern, to be able to extrapolate the remainder of the function. The work is an intersection of switching theory, computer science, and machine learning. Up until this point, we have not dealt with noisy data. Instead, it was assumed that all of the training data originally given was absolute truth. Furthermore, we have not yet developed any theory about what to do in cases that include noise. Therefore, the goal of this project was to explore some ideas about handling noise quickly in order to gain better insight into the problem. Moreover, this is an area in which I would personally like to pursue a dissertation topic. With the above stated, this project will explore dealing with noise in the domain of binary variables by preprocessing the training data. We will show quantitative results, draw conclusions, and point out future research directions. Machine Learning, Pattern Theory, Supervised Learning, C4.5 Noise.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1994
- Accession Number
- ADA286231
Entities
People
- Jeffrey A. Goldman
Organizations
- Wright Laboratory