Error-sensitive grading for model combination

Authors:
Surendra K. Singhi;Huan Liu
Affiliations:
Department of Computer Science and Engineering, Arizona State University, Tempe, AZ;Department of Computer Science and Engineering, Arizona State University, Tempe, AZ
Venue:
ECML'05 Proceedings of the 16th European conference on Machine Learning
Year:
2005

Citing 9
Cited 0

Original Contribution: Stacked generalization

Neural Networks
C4.5: programs for machine learning

C4.5: programs for machine learning
Bagging predictors

Machine Learning
Arbitrating among competing classifiers using learned referees

Knowledge and Information Systems
How to Make Stacking Better and Faster While Also Taking Care of an Unknown Weakness

ICML '02 Proceedings of the Nineteenth International Conference on Machine Learning
An Evaluation of Grading Classifiers

IDA '01 Proceedings of the 4th International Conference on Advances in Intelligent Data Analysis
Is Combining Classifiers with Stacking Better than Selecting the Best One?

Machine Learning
Data Mining: Practical Machine Learning Tools and Techniques, Second Edition (Morgan Kaufmann Series in Data Management Systems)

Data Mining: Practical Machine Learning Tools and Techniques, Second Edition (Morgan Kaufmann Series in Data Management Systems)
Issues in stacked generalization

Journal of Artificial Intelligence Research

Quantified Score

Hi-index	0.00

Visualization

Abstract

Ensemble learning is a powerful learning approach that combines multiple classifiers to improve prediction accuracy. An important decision while using an ensemble of classifiers is to decide upon a way of combining the prediction of its base classifiers. In this paper, we introduce a novel grading-based algorithm for model combination, which uses cost-sensitive learning in building a meta-learner. This method distinguishes between the grading error of classifying an incorrect prediction as correct, and the other-way-round, and tries to assign appropriate costs to the two types of error in order to improve performance. We study issues in error-sensitive grading, and then with extensive experiments show the empirically effectiveness of this new method in comparison with representative meta-classification techniques.