Supplementary Material for
On the Evaluation of Unsupervised Outlier Detection: Measures, Datasets, and an Empirical Study
by G. O. Campos, A. Zimek, J. Sander, R. J. G. B. Campello, B. Micenková, E. Schubert, I. Assent and M. E. Houle
Data Mining and Knowledge Discovery 30(4): 891-927, 2016, DOI: 10.1007/s10618-015-0444-8


This dataset has been preprocessed in different variants in the literature. We follow the procedure of Zhang et al. [1], using classes 1, 3, 4, 5, 6 and 7 as inliers and class 2 as outlier, selecting 1000 inliers vs. 13 outliers (class 2). The selection of instances is based on the test set. The processed dataset consists of 1013 instances represented in 9 attributes, with 13 outliers (1.28%) and 1000 inliers (98.72%).


[1] K. Zhang, M. Hutter, and H. Jin. A new local distance-based outlier detection approach for scattered real-world data. In Proc. PAKDD, pages 813-822, 2009.

Download all data set variants (328.2 kB). Access original data (shuttle.tst, [1] only uses test set)