Improving iForest with Relative Mass

Aryal, Sunil; Ting, Kai Ming; Wells, Jonathan R.; Washio, Takashi

doi:10.1007/978-3-319-06605-9_42

Cited by 38 publications

(18 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Local Outlier Probabilities (LoOP) employs the inverse ratio of probabilistic set distances [13], which can be viewed as a variant of density ratio. These relative scores have been shown to produce better anomaly detection performance [3,5] than their global counterparts such as k-th nearest neighbour distance [1,4,22] and path length [16,17]. Despite their popularity, our investigation reveals a key shortcoming of these ratio-based relative scores which has not been identified previously.…”

Section: Introductionmentioning

confidence: 66%

“…We compare the anomaly detection performance among NCAD, iForest [16], LOF [5], RMF [3] and LoOP [13]. A synthetic dataset and 12 benchmark datasets 1 are used in the evaluation.…”

Section: Methodsmentioning

confidence: 99%

“…A relative score is often defined as the aggregation of the ratios between the measurements of the query point and its neighbourhood, such as LOF [5], LoOP [13] and RMF [3]. A general form of relative score τ (x) can be defined as:…”

Section: A Key Shortcoming Of Relative Scoresmentioning

confidence: 99%

“…To detect both global and local anomalies, researchers have advocated the use of relative scores. For example, Local Outlying Factor (LOF) [5] computes a local reachability density ratio for each point based on a k-nearestneighbour density estimator; ReMass-iForest (RMF) [3] calculates the ratio of the numbers of points that fall into a leaf node and its immediate parent node in an isolation tree [16]. Local Outlier Probabilities (LoOP) employs the inverse ratio of probabilistic set distances [13], which can be viewed as a variant of density ratio.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Anomaly Detection via Neighbourhood Contrast

Chen

Ting

Chin

2020

Advances in Knowledge Discovery and Data Mining

Self Cite

View full text Add to dashboard Cite

Relative scores such as Local Outlying Factor and mass ratio have been shown to be better scores than global scores in detecting anomalies. While this is true, our analysis reveals for the first time that these relative scores have a key shortcoming: anomalies have greatly different relative scores if they are located in different regions where the curvatures of the density surface are very different. As a result, the lowscore anomalies could be ranked lower than some normal points. This revelation motivates (i) a new score called Neighbourhood Contrast (NC) which produces approximately the same high scores for all anomalies, regardless of varying curvatures of the density surface in different regions; and (ii) an anomaly detection method based on NC. Our experiments show that the proposed method which employs the new score significantly outperforms methods using the aforementioned relative scores on benchmark datasets.

show abstract

Section: Introductionmentioning

confidence: 66%

“…We compare the anomaly detection performance among NCAD, iForest [16], LOF [5], RMF [3] and LoOP [13]. A synthetic dataset and 12 benchmark datasets 1 are used in the evaluation.…”

Section: Methodsmentioning

confidence: 99%

Section: A Key Shortcoming Of Relative Scoresmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Anomaly Detection via Neighbourhood Contrast

Chen

Ting

Chin

2020

Advances in Knowledge Discovery and Data Mining

Self Cite

View full text Add to dashboard Cite

show abstract

“…The key to eliminate the outliers is to detect outliers, which is called the outlier detection and has wide application in the field of machine learning and research. An outlier is an observation which deviates so much from other observations as to arouse suspicions that it was generated by a different mechanism [24]. Isolation Forest (iForest) Algorithm [23] is an anomaly detector that measures without distance or density.…”

Section: Elimination Of Outliersmentioning

confidence: 99%

A Single-Way Ranging Localization of AUVs Based on PSO of Outliers Elimination

Fan

et al. 2018

Journal of Robotics

View full text Add to dashboard Cite

Localization of autonomous underwater vehicles (AUVs) is a very important and challenging task for the AUVs applications. In long baseline underwater acoustic localization networks, the accuracy of single-way range measurements is the key factor for the precision of localization of AUVs, whether it is based on the way of time of arrival (TOA), time difference of arrival (TDOA), or angle of arrival (AOA). The single-way range measurements do not depend on water quality and can be taken from long distances; however, there are some limitations which exist in these measurements, such as the disturbance of the unknown current velocity and the outliers caused by sensors and errors of algorithm. To deal with these problems, an AUV self-localization algorithm based on particle swarm optimization (PSO) of outliers elimination is proposed, which improves the performance of angle of arrival (AOA) localization algorithm by taking account of effects of the current on the positioning accuracy and eliminating possible outliers during the localization process. Some simulation experiments are carried out to illustrate the performance of the proposed method compared with another localization algorithm.

show abstract