getLocOut: Get Local Outlyingness

In tortnertuwien/lop: Local Projection Based Data Analysis

Description

Calculates the local description (getLocalDescription) for each observation of a data set. getLocalDescription is computed for each observation of data. To increase performance the package parrallel is used.

Usage

getLocOut(data, k, alpha=0.5)

Arguments

data	A data set to be evaluated for outliers. A matrix or a data frame may be handed over.
k	The number of observations used to define the environment by the k nearest neighbors. Only [k*alpha] of those observations are used for the definition of the projection space and for the covariance estimation used for the calculation of core distances.
alpha	A robustness parameter to adjust the number of oberservations used for the definition of the projection space. The optimal alpha depends on the expected number of outliers. Unless k is very small (<10), or no additional information is available, alpha should be kept at 0.5.

Value

ods	Matrix of orthogonal distances. The columns refere to the projection, the rows to the observation.
cds	Matrix of core distances. The columns refere to the projection, the rows to the observation.
ecds	Matrix of Euclidean core distances. The columns refere to the projection, the rows to the observation.
core	Matrix, documenting the [k*alpha] observations used for projections (columns). 1 referes to used, 0 to not used.
knn	Matrix, documenting the k nearest neighbors for each observation under evaluation. 1 referes to the knn, 0 to not not part of knn.

Author(s)

Thomas Ortner ([e-mail])

Examples

library(locout)
LocOut <- getLocOut(glass$data, k=20, alpha=0.5)

tortnertuwien/lop documentation built on May 30, 2019, 8:27 a.m.