Predicting Algorithm Accuracy with a Small Set of Effective Meta-Features

Author

Lee, Jun Won ; Carrier, Christophe Giraud

Author_Institution

Dept. of Comput. Sci., Brigham Young Univ., Provo, UT

fYear

2008

fDate

11-13 Dec. 2008

Firstpage

808

Lastpage

812

Abstract

We revisit 26 meta-features typically used in the context of meta-learning for model selection. Using visual analysis and computational complexity considerations, we find 4 meta-features whose values are directly relevant to certain ranges of predictive accuracy for 7 learning algorithms on 135 UCI datasets. Discretization of these 4 meta-features based on thresholds derived from our analysis significantly boosts the accuracy of the meta-level classification task.

Keywords

computational complexity; learning (artificial intelligence); pattern classification; UCI datasets; algorithm accuracy; computational complexity; learning algorithms; meta-features; meta-learning; meta-level classification task; model selection; visual analysis; Accuracy; Algorithm design and analysis; Application software; Clustering algorithms; Computational complexity; Data analysis; Machine learning; Machine learning algorithms; Prediction algorithms; Spatial databases; Accuracy Prediction; Meta-features; Meta-learning; Visual Analysis;

fLanguage

English

Publisher

ieee

Conference_Titel

Machine Learning and Applications, 2008. ICMLA '08. Seventh International Conference on

Conference_Location

San Diego, CA

Print_ISBN

978-0-7695-3495-4

Type

conf

DOI

10.1109/ICMLA.2008.62

Filename

4725071