Outlier Selection and One-Class Classification

What is common in a terrorist attack, a forged painting, and a rotten apple? The answer is: all three are anomalies; they are real-world observations that deviate from what is considered to be normal. Detecting anomalies is of utmost importance because an undetected anomaly can be dangerous or expensive. A human domain expert may suffer from three cognitive limitations: fatigue, information overload, and emotional bias. The cognitive limitations will hamper the detection of anomalies. Outlier-selection and one-class classification algorithms are capable of automatically classifying data points as outliers in large amounts of data. In this thesis we study to what extent outlier-selection and one-class classification algorithms can support domain experts with real-world anomaly detection.

The PDF can also be downloaded separately (228 pages, 10MB).

Figures

The figures in the thesis are created using Python, MATLAB and TikZ. The TikZ code of the figures can be found in /figures/tikz. To compile all the figures to PDF, I wrote a script called tikz2pdf.

$ tikz2pdf figures/tikz/*.tikz --template figures/thesis-template.tex --output figures/pdf/

Below are some figures from the thesis. Please note that these are rendered with a different font. Also, the conversion from PDF to PNG with ImageMagick isn't all that great.

figures/tikz/bg-banana-roc.tikz

figures/tikz/bg-multiclass.tikz

figures/tikz/eval-boxplot-preprocessing-pca.tikz

figures/tikz/mlc-mapping-auc-overview.tikz

figures/tikz/sos-densities.tikz

figures/tikz/sos-graph-matlab-binding.tikz

figures/tikz/sos-graphs-sample.tikz

figures/tikz/sos-nemenyi.tikz