A RNN Novel Approach for Unsupervised Distance-Based Outlier Detection
Journal Title: International Journal for Research in Applied Science and Engineering Technology (IJRASET) - Year 2016, Vol 4, Issue 7
Abstract
Detection of outliers in data defined as finding patterns in data that do not conform to normal behavior or data that do not conformed to expected behavior, such a data are called as outliers, anomalies, exceptions. Anomaly and Outlier have similar meaning. The analysts have strong interest in outliers because they may represent critical and actionable information in various domains, such as intrusion detection, fraud detection, and medical and health diagnosis. An Outlier is an observation in data instances which is different from the others in dataset. There are many reasons due to outliers arise like poor data quality, malfunctioning of equipment, ex credit card fraud. Data Labels associated with data instances shows whether that instance belongs to normal data or anomalous. Based on the availability of labels for data instance, the anomaly detection techniques operate in one of the three modes are 1)Supervised Anomaly Detection, techniques trained in supervised mode consider that the availability of labeled instances for normal as well as anomaly classes in a a training dataset. 2) Semi-supervised Anomaly Detection, techniques trained in supervised mode consider that the availability of labeled instances for normal, do not require labels for the anomaly class. 3) Unsupervised Anomaly Detection, techniques that operate in unsupervised mode do not require training data. There are various methods for outlier detection based on nearest neighbors, which consider that outliers appear far from their nearest neighbors. Such methods base on a distance or similarity measure to search the neighbors, with Euclidean distance. Many neighbor-based methods include defining the outlier score of a point as the distance to its kthnearest neighbor (k-NN method), some methods that determine the score of a point according to its relative density, since the distance to the kth nearest neighbor for a given data point can be viewed as an estimate of the inverse density around it.
Authors and Affiliations
M. Siva Kumar, G. Prasadbabu
Keywords
Related Articles
- EP ID EP22441
- DOI -
- Views 226
- Downloads 3
A Graphical user interface for web crawler
The number of web pages is increasing into millions and trillions around the world. To make searching much easier for users, web search engines came into existence. Web Search engines are used to find specific informati...
An Investigation of Heat Transfer in Vertical Shaft Kiln Process of Cement Manufacture: Part I - Convection
Packed bed heat transfer associated with the black meal vertical shaft kiln (VSK) process of cement manufacture was studied using varying feed sizes of the raw material and air flow rates using an experimental prototype...
Comparison Of Various 32-Bit Parallel Prefix Adders
In Very Large Scale Integration (VLSI) designs, Parallel prefix adders (PPA) have the higher delay performance. This paper investigates four varieties of PPA’s (Kogge Stone Adder (KSA), Spanning Tree Adder (STA), Brent...
Optimization of Surface Roughness on High Carbon High Chromium Steel with Graphite Tool
Electrical discharge machining (EDM) is a nontraditional machining process and it is used to machine hard material components, that are precise and difficult-to-machine such as heat treated tool steels. In die Sinking E...
A Three Level Ac-Dc Conversion with Boost Converter
A new three level AC-DC conversion along with boost converter is proposed. The proposed converter integrates the operation of the boost power factor correction and the three level dc-dc converters. The converter is made...