Combination fairness with scores in outlier detection ensembles

Article Computer Science, Information Systems

Outlier detection in social networks leveraging community structure

Arnab Dey et al.

Summary: Social networks have become essential in today's world and are increasingly used for communication worldwide. The large amount of data transferred over social networks necessitates the need for security precautions. This paper introduces a new technique that detects anomalies in a network from a global perspective using the network community structure, outperforming existing algorithms in terms of accuracy and speed.

INFORMATION SCIENCES (2023)

添加到收藏夹

Article Computer Science, Information Systems

Incomplete mixed data-driven outlier detection based on local-global neighborhood information

Ran Li et al.

Summary: Outlier detection is a challenging task due to the nature of ubiquitous, incomplete, redundant, noisy, and mixed data. To address this challenge, this paper proposes an ILGNI network that considers both local and global information from incomplete mixed data. The network enhances connectivity between similar objects and weakens connectivity between heterogeneous objects, allowing for efficient graph-based outlier detection. Experiments on telecom fraud datasets demonstrate that the proposed algorithm achieves enhanced outlier detection performance with low time complexity and is applicable to various types of datasets.

INFORMATION SCIENCES (2023)

添加到收藏夹

Article Computer Science, Artificial Intelligence

An individual fairness based outlier detection ensemble *

Gargi Mishra et al.

Summary: Outlier detection ensembles are usually better than individual detectors, but they may still have biases towards certain groups or individuals. To address this, the notion of fairness is introduced. This paper proposes an individual fairness-based approach (iFselect) for selecting members in an outlier detection ensemble. The method considers the relative similarity of instances and their outlier scores to achieve a balance between performance and fairness. Experiments show the superiority of iFselect in fairness and comparable overall performance to state-of-the-art methods.

PATTERN RECOGNITION LETTERS (2023)

添加到收藏夹

Article Computer Science, Information Systems

Unsupervised anomaly detection ensembles using item response theory

Sevvandi Kandanaarachchi

Summary: Ensemble learning combines multiple algorithms or models to improve predictive performance and is used in various fields. This study introduces the use of Item Response Theory (IRT), a model used in educational psychometrics, to construct an unsupervised anomaly detection ensemble. The effectiveness of the IRT ensemble is demonstrated and shown to outperform other ensemble techniques, even with low correlation values among the anomaly detection methods.

INFORMATION SCIENCES (2022)

添加到收藏夹

Article Computer Science, Information Systems

Graph autoencoder-based unsupervised outlier detection

Xusheng Du et al.

Summary: Outlier detection technologies are crucial in various domains. The study proposes a novel graph neural network structure called the graph autoencoder (GAE) for outlier detection in Euclidean structured data. Experimental results show that GAE achieves the highest ROC AUC among multiple datasets.

INFORMATION SCIENCES (2022)

添加到收藏夹

Article Computer Science, Information Systems

FairLOF: Fairness in Outlier Detection

P. Deepak et al.

Summary: This paper focuses on fair outlier detection over multiple multi-valued sensitive attributes and proposes FairLOF, a fair outlier detection method inspired by the popular LOF formulation. Through empirical evaluation, FairLOF is shown to significantly improve fairness with marginal degradations in result quality compared to the fairness-agnostic LOF method. Additionally, a generalization of FairLOF, named FairLOF-Flex, offers possibilities for further deepening fairness in outlier detection.

DATA SCIENCE AND ENGINEERING (2021)

添加到收藏夹

Article Computer Science, Artificial Intelligence

Building outlier detection ensembles by selective parameterization of heterogeneous methods☆

Akanksha Mukhriya et al.

Summary: AnD-SELECT is an accurate and diverse selector for selecting members in unsupervised outlier detection, aiming to choose accurate parameter variants while avoiding selection of both ends.

PATTERN RECOGNITION LETTERS (2021)

添加到收藏夹

Article Computer Science, Artificial Intelligence