Funded by the European Union. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the European Education and Culture Executive Agency (EACEA). Neither the European Union nor EACEA can be held responsible for them.


On the Evaluation of Outlier Detection and One-Class Classification Methods

This paper provides an overall comparison between one-class classification algorithms and unsupervised outlier detection methods extended to the one-class classification scenario. The proposed methods' performances have been compared in terms of ROC AUC and [email protected] The Model Selection has been performed for each method through a 10 fold cross-validation procedure on the 80% of the data available. The remaining 20% has been exploited for testing purposes. The experiments have been conducted on 31 real-world datasets from the UCI Machine Learning Repository. SVDD and kNNglobal turned out to be the best performing algorithms for one-class classification scenarios.

Scientific Paper

Machine Learning

Target Group:


Cite as:
L. Swersky, H. O. Marques, J. Sander, R. J. G. B. Campello and A. Zimek, "On the Evaluation of Outlier Detection and One-Class Classification Methods," 2016 IEEE International Conference on Data Science and Advanced Analytics (DSAA), 2016, pp. 1-10.

Author of the review:
Davide Ilardi
University of Genoa


You have to login to leave a comment. If you are not registered click here