A review on outlier detection techniques on data stream by using different approaches of K-Means algorithm

Abstract

Data Stream mining has gained attraction from many researchers as there is need to mine large dataset which pose different challenges for researchers. Stream data is different compared to normal data as they are continuously produced from different applications which impose different challenges like massive, infinite, concept drift for processing. An object that does not obey the behavior of normal data object is called outliers. Outlier detection is used in different applications like fraud detection, intrusion detection, track environmental changes, medical diagnosis so there is need to detect outliers from data streams. Various approaches are used for outlier detection. Some of them use K-Means algorithm for outlier detection in data streams which help to create a similar group or cluster of data points. Data stream clustering techniques are highly helpful to cluster similar data items in data streams and also to detect the outliers from them, so they are called cluster based outlier detection. K-means algorithm is partition based algorithm which is used for clustering datasets into number of clusters. It is most common and popular algorithm for clustering due to its simplicity and efficiency. Purpose of this paper is to review of different approaches of outlier detection which is used for K-Means algorithm for clustering dataset with some other methods. Different application areas of outlier detection are discussed in this paper.