Neighborhood representative for improving outlier detectors

Yang, Jiawei; Chen, Yu; Rahardja, Sylwan

doi:10.1016/j.ins.2022.12.041

Cited by 18 publications

(3 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, Ke et al [36] proposed a method called group similarity system (GSS) for unsupervised outlier detection and Yang et al [37] proposed a data pre-processing technique called neighborhood representative (NR) to detect collective outliers using exiting outlier detectors. GSS partitions the data into non-overlapped groups and judges the groups as outliers by considering the mean of the outlier scores of the objects in each group.…”

Section: Discussionmentioning

confidence: 99%

Smoothing Outlier Scores Is All You Need to Improve Outlier Detectors

Yang,

Rahardja,

Fränti

2024

IEEE Trans. Knowl. Data Eng.

View full text Add to dashboard Cite

We hypothesize that similar objects should have similar outlier scores. To the best of our knowledge, all existing outlier detectors calculate the outlier score for each object independently regardless of the outlier scores of the other objects. Therefore, they do not guarantee that similar objects have similar outlier scores. To verify our proposed hypothesis, we propose an outlier score postprocessing technique for outlier detectors, called neighborhood averaging (NA) for neighborhood smoothing in outlier score space. It pays attention to objects and their neighbors and guarantees them to have more similar outlier scores than their original scores. Given an object and its outlier score from any outlier detector, NA modifies its outlier score by combining it with its k nearest neighbors' scores. We demonstrate the effectivity of NA by using the well-known k nearest neighbors (k-NN). Experimental results show that NA improves all 10 tested baseline detectors by 13% on average relative to the original results (from 0.70 to 0.79 AUC) evaluated on nine real-world datasets. Moreover, deep-learning-based detectors and even outlier detectors that are already based on k-NN are also improved. The experiments also show that in some applications, the choice of detector is no more significant when detectors are jointly used with NA. This may pose a challenge to the generally considered idea that the data model is the most important factor. We open our code on www.outlierNet.com for reproducibility.

show abstract

Section: Discussionmentioning

confidence: 99%

Smoothing Outlier Scores Is All You Need to Improve Outlier Detectors

Yang,

Rahardja,

Fränti

2024

IEEE Trans. Knowl. Data Eng.

View full text Add to dashboard Cite

show abstract

“…Based on KDE, Zhang et al [21] and Dong et al [20] revamped the calculation of the outlier score in studies. Yang et al [32] designed a framework named neighborhood averaging to postprocess the outlier scores in existing algorithms like LOF, using the average score of neighbors as the new score of objects, based on the hypothesis that similar objects should have similar outlier scores, and demonstrated the effectivity.…”

Section: Related Workmentioning

confidence: 99%

“…Inspired by the above methods, we utilize the PCA method and natural neighbor search algorithm as preprocessing steps of our proposed LDF method. In addition, based on the idea of neighborhood averaging [32], we propose the concept of neighborhood similarity in this paper. However, unlike the neighborhood averaging works on final scores and is just simple averaging of scores, neighborhood similarity works on density and can characterize the outlier properties in a more refined way.…”

Section: Related Workmentioning

confidence: 99%

An Outlier Detection Algorithm Based on Local Density Feedback Outlier Factor

Zhang,

Hou,

Jia

et al. 2024

Preprint

View full text Add to dashboard Cite

Neighborhood similarity assumes that adjacent objects should exhibit similar characteristics. To address the issue that previous density-based algorithms neglected neighborhood similarity and global information, resulting in low accuracy for outlier detection, we propose an outlier detection algorithm based on local density feedback outlier factor. Firstly, principal component analysis method is employed combining with natural neighbor search algorithm for density estimation. Subsequently, a density feedback mechanism is introduced for post-processing density values, which defines the density differences within the neighborhood as feedback signals, and update density feedback values for each object based on the principle of neighborhood similarity. Through multiple iterations of feedback, individual objects can aggregate global information. Finally, feedback outlier factor is defined to quantify outlier degrees. Experimental validation demonstrates the effectiveness of the proposed algorithm in enhancing outlier detection performance.

show abstract