Func.trans

raw.score

should be a character specifying  the method used to generate the raw score. It has 3 possible values, "ABOD", "SOD", and "FBOD".

method


This function calculate the transformed outlier scores, with the aim of unifying the results from different methods.
The method is based on the work of Kriegel, H.P., Kroger, P., Schubert, E., Zimek, A., Interpreting and unifying outlier scores, 2011.
It consists of two steps, regularization and normalization.
For the ABOD scores, logarithmic inversion is used for regularization
For the SOD scores, no action is taken to perform regularization
For the FBOD method, the basic regularization, i.e., score-1, is used for regularization
For the normalization step, the gaussian scaling method is used.
The final output can be interpreted as the outlier probability, ranging from 0 to 1.


Three high-dimensional outlier detection algorithms and a outlier unification scheme are implemented in this package. The angle-based outlier detection (ABOD) algorithm is based on the work of Kriegel, Schubert, and Zimek [2008]. The subspace outlier detection (SOD) algorithm is based on the work of Kriegel, Kroger, Schubert, and Zimek [2009]. The feature bagging-based outlier detection (FBOD) algorithm is based on the work of Lazarevic and Kumar [2005]. The outlier unification scheme is based on the work of Kriegel, Kroger, Schubert, and Zimek [2011].

Func.trans: Outlier score transformation

Description

Usage

Arguments

Value