Novel outlier detection method based on clustering

Abstract

Data Mining presents the process of refining the unknown information from huge data. newlineAdvancement has led to a sudden upsurge in large number of algorithms that effectively tackle newlinethe regular and computing task of data mining. Data mining in other words is a process of newlineretrieving knowledge from large amount of data. Data mining offers great promise in helping newlineorganizations to uncover patterns and knowledge usually inconspicuous in their data that can newlinebe used to analyze the relationship and performance of customers and products with future newlinetrends. The key idea of this article is to provide an overview of data mining algorithm. The newlineimplemented outcomes will provide useful and profitable results. The software helps to newlinediscover concealed facts and interesting knowledge, to provide help in decision making newlineprocess. Data mining is the approach which searches for new, valuable, and nontrivial newlineinformation from large volumes of data. The study includes the techniques that utilizes newlineclustering and are beneficial for pattern recognition. The study aims at providing the review of newlineclustering techniques and their applications in pattern recognition. The discussion on the study newlinewill guide the researchers for improving their research direction. newlineAt present outlier detection is an effective and functioning area of data mining. Data newlinemining can be outlined as gaining knowledge from large databases. Data mining is a boon to newlineindustries and organizations such as to identifying the patterns hidden in the data that can be newlinefurther analysed. The generated patterns help to understand the relationship behaviour with newlineproducts and the future trends. One of the analysis can be detection of Outliers. Outliers can be newlinedefined as the data point those substantially unlike from other data points in the data set. newlineExistence of Outliers can change the results. Outliers apply many techniques of data mining newlinelike classification and clustering. The prime idea of this paper is to provide an overview of newlineoutliers and various approaches for outlier detection. In this era detection of Outlier is a newlinesignificant area in the field of data mining. Outliers can be evolved from multiple sources like newlineclerical error, system error, mechanical faults, or may be generated during capturing of data newlinefrom different sources. It forms a pivotal point in data mining that can be searched by using newlinesupervised or unsupervised learning techniques. It can detect outlier generated along with newlinerequired information that creates noise in the data. In this thesis we intend to present a newlinecomparative study between distance based and angle based outlier detection methods over data newlinesets for outlier detection. Distance based concept uses some distance methods like Euclidean newlinedistance or Manhattan distance. It not only requires the understanding of mathematicalproperties but also the relevant knowledge of the data. Angle based approach emphasizes on newlinethe deviation of angles between two data points and it is a parameter free approach. Further a newlinenew approach has been introduced which is an integrated method of angle and distance based newlineapproaches. This technique would be used to detect outliers in the given set of data. This study newlineconcurrently also aims at providing the review of clustering technique and various outlier newlineanalysis techniques in the data and to use this comparison for further research studies.