Certain investigations on performance analysis of diverse outlier detection schemes for various immense applications using various methods

Abstract

Data mining is the field to study and analyze the process of discovering the knowledge in terms of useful patterns, outliers and co-relation etc., Most of the time outliers are ignored by considering them as a noise. When looking deeper, outliers are having potential significance on the data. Ignoring outliers by simply considering as a noise is not the right decision in all the scenarios. Outliers are taking a vital role in the applications like fraud detection, intrusion detection and quality control etc. In a nutshell, outlier detection is the systematic approach or algorithematic practice of identifying outliers from different kinds of datasets. The method and the way of detecting outliers vary according to the application domain. Studying and analyzing outliers enables the visibility of hidden knowledge in the data. newlineIn this thesis, the general problem of outlier detection is studied based on the various applications. A detailed survey is presented which is giving a comprehensive study of outlier detection, diverse new models developed for outlier detection in respect of contemporary applications. Finding outliers in distributed repository or dataset is always a challengeable task. In scattered repository the data is physically distributed and stored geographically in different locations. The method of unit concentration scheme for outlier detection used to detect outlier in the scattered repositories with information safety and without data seepages. newline