Tutorial 2 (morning)
Evolutionary Computation Approaches to Mining Biomedical Data
John Holmes

• Overview
• Detailed Content
• Intended Audience
• Pre-requisite Knowledge
• Important Dates
• Presenter

Overview

The interest in using EC tools in KDD has been growing exponentially over the past several years, as evidenced by the burgeoning of KDD-related papers and presentations in EC-related literature and meetings.

Detailed Content

• Introduction
• Survey of tools for knowledge discovery
• Characteristics of biomedical data
• Missing data
• Feature selection issues
• Preparing biomedical data for KDD
• Introduction to evolutionary computation (EC): Each of the following will include knowledge representation, functional analysis (via pseudocode), and comparative strengths and weaknesses, and will be demonstrated using a simple biomedical dataset.
• Genetic algorithms
• Learning classifier systems
• Genetic programming
• Evolutionary algorithms
• Applications of EC to KDD: Each application will use a simple biomedical dataset for demonstration
• Rule discovery
• Emergence of clinical prediction rules
• Clustering
• Evaluation
• Metrics
• Classification and prediction accuracy
• Sensitivity, specificity, and the area under the receiver operating characteristic curve
• Predictive value
• Likelihood ratios
• Validation procedures
• Applying the metrics to validation
• Methods for comparing EC tools with others used in KDD
• Validation datasets
• Expert panels
• Summary

Intended Audience

This tutorial is intended for those with some introductory knowledge of the KDD process and data mining tools. It will draw attendees from two separate, yet increasingly cross-disciplinary, domains: data miners with an interest in new, EC-based tools, as well as EC researchers with an interest in KDD applications

Pre-requisite Knowledge

No prior experience with evolutionary computation (EC) is required, although those who are familiar with this area would benefit by seeing how familiar EC techniques can be applied to mining biomedical data.

Important Dates

Presenter

John H. Holmes, PhD, is an internationally recognized expert in applying evolutionary computation methods to knowledge discovery in biomedical databases. He is a regular contributor to GECCO (the Genetic and Evolutionary Computation Conference), a reviewer for Artificial Intelligence in Medicine, Evolutionary Computation, IEEE Transactions on Evolutionary Computation, among several clinical journals, specializing in KDD issues and applications. Dr. Holmes has also co-authored a chapter on using learning classifier systems in knowledge discovery (Bull L (ed): Applications of Learning Classifier Systems Berlin:Springer 2004, 15-67). He has given numerous tutorials and lectures on biomedical KDD at such venues as the Fall Symposium of the American Medical Informatics Association, Medinfo, The Drug Information Association, and the Centers for Disease Control and Prevention, as well as universities in the United States and Canada. His tutorials have been very well attended; most recently, his KDD tutorial at Medinfo 2004 drew over 50 people, one of the largest tutorial audiences at that meeting.