A New Bootstrapping Method to Improve Classification Performance in Learning Classifier Systems

Authors:
John H. Holmes;Dennis R. Durbin;Flaura K. Winston
Affiliations:
-;-;-
Venue:
PPSN VI Proceedings of the 6th International Conference on Parallel Problem Solving from Nature
Year:
2000

Citing 9
Cited 3

Induction: processes of inference, learning, and discovery

Induction: processes of inference, learning, and discovery
NEWBOOLE: a fast GBML system

Proceedings of the seventh international conference (1990) on Machine learning
C4.5: programs for machine learning

C4.5: programs for machine learning
Predictive data mining: a practical guide

Predictive data mining: a practical guide
Genetic Algorithms in Search, Optimization and Machine Learning

Genetic Algorithms in Search, Optimization and Machine Learning
An Empirical Comparison of Voting Classification Algorithms: Bagging, Boosting, and Variants

Machine Learning
A Tale of Two Classifier Systems

Machine Learning
Query Learning Strategies Using Boosting and Bagging

ICML '98 Proceedings of the Fifteenth International Conference on Machine Learning
Boosting a Strong Learner: Evidence Against the Minimum Margin

ICML '99 Proceedings of the Sixteenth International Conference on Machine Learning

Applying a Learning Classifier System to Mining Explanatory and Predictive Models from a Large Clinical Database

IWLCS '00 Revised Papers from the Third International Workshop on Advances in Learning Classifier Systems
A Bigger Learning Classifier Systems Bibliography

IWLCS '00 Revised Papers from the Third International Workshop on Advances in Learning Classifier Systems
A Representation for Accuracy-Based Assessment of Classifier System Prediction Performance

IWLCS '01 Revised Papers from the 4th International Workshop on Advances in Learning Classifier Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

A new technique for improving the classification performance of learning classifier systems (LCS) was developed and applied to a real-world data mining problem. EpiCS, a stimulus-response LCS, was adapted to perform prevalence-based bootstrapping, wherein data from training and testing sets were sampled according to the prevalence of the individual classes, rather than randomly using the class distribution inherent in the data. Prevalence-based bootstrapping was shown to improve classification performance significantly on training and testing. Furthermore, this procedure was shown to enhance EpiCS's classification performance on testing compared to a well-known decision tree inducer (C4.5) when similar bootstrapping procedures were applied to the latter.