An evaluation of data stream clustering algorithms

Data stream clustering is a hot research area due to the abundance of data streams collected nowadays and the need for understanding and acting upon such sort of data. Unsupervised learning (clustering) comprises one of the most popular data mining tasks for gaining insights into the data. Clustering is a challenging task, while clustering over data streams involves additional challenges such as the single pass constraint over the raw data and the need for fast response. Moreover, dealing with an infinite and fast changing data stream implies that the clustering model extracted upon such sort of data is also subject to evolution over time. Several stream clustering surveys exist already in the literature; however, they focus on a theoretical presentation of the surveyed algorithms. On the contrary, in this paper, we survey the state‐of‐the‐art stream clustering algorithms and we evaluate their performance in different data sets and for different parameter settings.