Authors:
David Štefka
and
Martin Holeňa
Affiliation:
Institute of Computer Science, Academy of Sciences of the Czech Republic, v.v.i., Czech Republic
Keyword(s):
Classifier aggregation, Classifier combining, Classification confidence.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Biomedical Engineering
;
Biomedical Signal Processing
;
Computational Intelligence
;
Data Manipulation
;
Evolutionary Computing
;
Health Engineering and Technology Applications
;
Human-Computer Interaction
;
Knowledge Discovery and Information Retrieval
;
Knowledge-Based Systems
;
Machine Learning
;
Methodologies and Methods
;
Neurocomputing
;
Neurotechnology, Electronics and Informatics
;
Pattern Recognition
;
Physiological Computing Systems
;
Sensor Networks
;
Soft Computing
;
Symbolic Systems
Abstract:
Classifier aggregation is a method for improving quality of classification. Instead of using just one classifier, a team of classifiers is created, and the outputs of the individual classifiers are aggregated into the final prediction. In this paper, we study the potential of using measures of local classification confidence in classifier aggregation methods. We introduce four measures of local classification confidence and study their suitability for classifier aggregation. We develop two novel classifier aggregation methods which utilize local classification confidence and we compare them to two commonly used methods for classifier aggregation. The results on four artificial and five real-world benchmark datasets show that by incorporating local classification confidence into classifier aggregation methods, significant improvement in classification quality can be obtained.