Author:Mark Hall
Category:Outlier
Changes:Fixed a capabilities bug in the filter that prevented running on datasets with a unary class set. The LOF classifier no longer produces probabilities that are exactly zero, as the Evaluation class will interpret an all zero distribution as unclassified. Also reversed the order/meaning of the probabilities output - it now outputs 1 - normalized LOF score for the first probability (i.e. probability of 'normal') and the normalized LOF score for the second (outlier) class (in the binary class case). This makes more sense in the unary case where training data typically contains instances of the normal class.
Date:2015-12-08
Depends:weka (>=3.7.2)
Description:A filter that applies the LOF (Local Outlier Factor) algorithm to compute an "outlier" score for each instance in the data. Can use multiple cores/cpus to speed up the LOF computation for large datasets. Nearest neighbor search methods and distance functions are pluggable. Also contains a classifier that uses LOF. The class attribute (ignored at training time) is expected to have two values, where the second is for "normal" examples. If normal and outlier instances are known in the data then this classifier allows LOF to be evaluated using Weka's evaluation routines for classifiers.

For more information, see:

Markus M. Breunig, Hans-Peter Kriegel, Raymond T. Ng, Jorg Sander (2000). LOF: Identifying Density-Based Local Outliers. ACM SIGMOD Record. 29(2):93-104.
License:GPL 3.0
Maintainer:Weka team <wekalist{[at]}list.scms.waikato.ac.nz>
PackageURL:http://prdownloads.sourceforge.net/weka/localOutlierFactor1.0.6.zip?download
URL:http://weka.sourceforge.net/doc.packages/localOutlierFactor
Version:1.0.6