Survey: Clustering Techniques of Data Stream

Abstract

Data stream mining is an emerging specialty in the field of mining huge data that extract useful knowledge of the whole of the data stream. There are several mining processes to handle the data stream. One of the most important and widely used is the clustering data stream. The clustering is either hard (exclusive) or fuzzy (soft) clustering. Recently, significant sources are made available to generate data stream, therefore, the clustering of these data is an important and vital topic for many researchers. Several data stream algorithms have been proposed by researchers during the past years, while some have developed other algorithms.The data stream clustering varies from the traditional clustering in many principles where the main differences between them are explained. Meanwhile, data stream clustering has new challenges such as the single pass on the raw data sets, the unbounded size of this data and the high speed arriving of data samples. But the most prominent one is the dynamic nature of the data.This paper presents a comprehensive study on the hard data stream clustering methods and their algorithms. In addition to the advantages and disadvantages of these methods. Where the paper deals with many aspects that surrounds the data stream, such as stream conditions, challenges, dynamics that it needs, and it changes over time. Then, it presented a transition to the modern trends of clustering algorithms and their utility in online applications. The survey aims to be an auxiliary reference for the researcher in determining the clustering algorithm that compatible with the available data set to achieve the desired goal.