Tensor-based anomaly detection: An interdisciplinary survey

Abstract

Traditional spectral-based methods such as PCA are popular for anomaly detection in a variety of problems and domains. However, if data includes tensor (multiway) structure (e.g. space-time-measurements), some meaningful anomalies may remain invisible with these methods. Although tensor-based anomaly detection (TAD) has been applied within a variety of disciplines over the last twenty years, it is not yet recognized as a formal category in anomaly detection. This survey aims to highlight the potential of tensor-based techniques as a novel approach for detection and identification of abnormalities and failures. We survey the interdisciplinary works in which TAD is reported and characterize the learning strategies, methods and applications; extract the important open issues in TAD and provide the corresponding existing solutions according to the state-of-the-art.

Anomaly detection in spatiotemporal data via regularized non-negative tensor analysis

“…Recently, Hadi et al did a comprehensive review of the tensor-based anomaly detection techniques, most of which were developed beyond the scope of computer science (Fanaee-T and Gama 2016a). Here we focus on those tensor based methods that are most related to our work.…”

“…Most of the semi-supervised models are designed for real-time anomaly detection and can be divided into two categories (Fanaee-T and Gama 2016a). Methods in both categories use normal data (i.e., positive samples) to construct a tensor and use the decomposition results as a baseline.…”

“…Matrix-based methods such as PCA, though proved to be powerful in many areas, cannot deal with multi-way data. Some empirical evidence (Fanaee-T and Gama 2016a) has shown the superiority of tensor-based methods over traditional matrix-based methods. On this occasion, tensor-based methods are appropriate since the tensor could store the natural multi-dimensional relationship of the data.…”

“…On this occasion, tensor-based methods are appropriate since the tensor could store the natural multi-dimensional relationship of the data. Tensor-based anomaly detection methods (Fanaee-T and Gama 2016a) are able to decompose a multi-way tensor into information factors and reveal their relationships based on probability models. However, none of the existing tensor-based techniques are specifically developed to detect anomalous traffic patterns, such as sudden increase in the volume of incoming or outgoing people of a particular region or irregular traffic flow among several regions.…”

A high-order representation and classification method for transcription factor binding sites recognition in Escherichia coli

“…A promising way to deal with this issue is to use multidimensional array-based representation [29,27]. Actually, this type of representation has been successfully applied to EEG signals classification in biomedical engineering [26,25,7], image processing in computer vision or pattern recognition [47,44,21], and other fields [40,33,12].…”