A Survey on Anomalies Detection Techniques and Measurement Methods

Muruti, Gopinath; Rahim, Fiza Abdul; Ibrahim, Zul-Azri

doi:10.1109/ains.2018.8631436

Cited by 28 publications

(20 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [9], [10], the existing techniques for anomaly detection which include statistical, neural network based, and other ML based techniques are discussed. Various books [11]- [14] also discussed the techniques for anomaly detection.…”

Section: A Related Workmentioning

confidence: 99%

Anomalous Example Detection in Deep Learning: A Survey

et al. 2020

View full text Add to dashboard Cite

Deep Learning (DL) is vulnerable to out-of-distribution and adversarial examples resulting in incorrect outputs. To make DL more robust, several posthoc (or runtime) anomaly detection techniques to detect (and discard) these anomalous samples have been proposed in the recent past. This survey tries to provide a structured and comprehensive overview of the research on anomaly detection for DL based applications. We provide a taxonomy for existing techniques based on their underlying assumptions and adopted approaches. We discuss various techniques in each of the categories and provide the relative strengths and weaknesses of the approaches. Our goal in this survey is to provide an easier yet better understanding of the techniques belonging to different categories in which research has been done on this topic. Finally, we highlight the unsolved research challenges while applying anomaly detection techniques in DL systems and present some high-impact future research directions.

show abstract

Section: A Related Workmentioning

confidence: 99%

Anomalous Example Detection in Deep Learning: A Survey

et al. 2020

View full text Add to dashboard Cite

show abstract

“…For this, in the matrix every elements of negative value are assigned zero. If in a row of matrix, every elements are zero, they must be re-calculated by sequence of arbitrary numbers in the interval [0, 1] and without violating the constraints, the matrix experiences the resulting changes: (6) The FPSO algorithm like former algorithms, requires a function for estimating the general results called fitness function. In this work "equation.…”

Section: (4) (5)mentioning

confidence: 99%

“…In this work "equation. (6)" is used for evaluating the solutions. 7where K represents a constant and J m denotes objective function.…”

Section: (4) (5)mentioning

confidence: 99%

“…To prevent fraud events there are generally two procedures: Classification method and Abnormality identification. In the first method, a classifier is trained with the given patterns of normal and fraud transactions using supervised procedures [6].Second method is having the ability to filter the new deals that is uneven against the profile of the card holder by computing the distance of the inward transactions with the profile [5]. The two above methods have its own drawbacks.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Fuzzy Particle Swarm Optimization Feature Selection and Aggrandized Classifier for Uncovering Frauds in Credit Card Deals

Kumar

2020

IJEAT

View full text Add to dashboard Cite

In today's economy, credit card plays a very important role. The rise of credit card customers improved, credit card scam cases were also on the rise. Numerous procedures are anticipated to challenge the evolution of the frauds in credit cards. In this research work, proposed an innovative fraud detection method which utilizes the similar cardholder’s behavioral patterns to construct a current cardholder’s interactive profile in order to stay away from the credit card scams. However, the selection of optimal features from the samples and the decision cost for accuracy becomes main important problem. To illuminate these issues this proposed research work presents an innovative fraud detection technique that makes out of four phases: 1. To augment a cardholder’s behavioral styles, first we divide all cardholders into distinctive groups making use of the cardholder’s historical transaction data such that the members of each group have the similar transaction behavior by K-means. 2. Introduces a new Fuzzy Particle Swarm Optimization (FPSO) feature selection for the amplification of fraud detection in credit cards. 3. By means of a prolonged wrapper method, an ensemble classification are performed by Aggrandized Kernel based Support Vector Machine (AKSVM).4.Refreshing the cardholder’s social profile with an input system. This Proposed work adopts the external quality metrics as Accuracy, Recall, Concept drift rate and Fraud feature rate. The UCI dataset is used and is done in MATLAB framework. The analytical measures were used to estimate the routine of the mentioned fraud detection technique. The simulation results show that this proposed innovative fraud detection method provides better accuracy results than other fraud detection techniques. The low concept drift rate results the gain of the innovative method to classify the transactions accurately.

show abstract

“…Furthermore, the anomaly detection for time and position of the suspected feeders is in the lack, which is also important information for the maintenance staff. As one of the statistical analysis methods to detect the abnormality, the theory of the control chart is mature, which can provide a statistically valid answer for anomaly detection [18]. The control chart generated from the historical data can analyze and judge the steady state of the operation process, which is suitable to identify the abnormal time of the suspected feeder.…”

Section: Introductionmentioning

confidence: 99%

A Data-Driven Combined Algorithm for Abnormal Power Loss Detection in the Distribution Network

Long

Chen

et al. 2020

IEEE Access

View full text Add to dashboard Cite

Power loss, consisting of technical loss (TL) and non-technical loss (NTL), reflects the effective utilization rate of energy and the management level of power grids. This paper proposes a data-driven combined algorithm to systematically identify anomalies of power loss in the distribution network, including the abnormal type, time, and position. The detection process contains three stages: abnormal feeder detection, abnormal time detection, and abnormal position detection. The suspected abnormal feeders are first detected from all feeders in the distribution network by the data-driven algorithm based on the daily power supply and electricity sales data. Then, the control chart is employed to further monitor the fluctuation of the power loss of each suspected abnormal feeder and discover its abnormal time. Based on the detected abnormal time, its abnormal position is finally located through the risk assessment technology. Numerous experiments based on the real data show that the proposed data-driven combined algorithm can effectively detect and analyze abnormal power loss in the distribution network.

show abstract

A Survey on Anomalies Detection Techniques and Measurement Methods

Cited by 28 publications

References 46 publications

Anomalous Example Detection in Deep Learning: A Survey

Anomalous Example Detection in Deep Learning: A Survey

Fuzzy Particle Swarm Optimization Feature Selection and Aggrandized Classifier for Uncovering Frauds in Credit Card Deals

A Data-Driven Combined Algorithm for Abnormal Power Loss Detection in the Distribution Network

Contact Info

Product

Resources

About