Dynamic Sparse Subspace Clustering for Evolving High-Dimensional Data Streams

Abstract

In an era of ubiquitous large-scale evolving data streams, data stream clustering (DSC) has received lots of attention because the scale of the data streams far exceeds the ability of expert human analysts. It has been observed that high-dimensional data are usually distributed in a union of low-dimensional subspaces. In this article, we propose a novel sparse representation-based DSC algorithm, called evolutionary dynamic sparse subspace clustering (EDSSC). It can cope with the time-varying nature of subspaces underlying the evolving data streams, such as subspace emergence, disappearance, and recurrence. The proposed EDSSC consists of two phases: 1) static learning and 2) online clustering. During the first phase, a data structure for storing the statistic summary of data streams, called EDSSC summary, is proposed which can better address the dilemma between the two conflicting goals: 1) saving more points for accuracy of subspace clustering (SC) and 2) discarding more points for the efficiency of DSC. By further proposing an algorithm to estimate the subspace number, the proposed EDSSC does not need to know the number of subspaces. In the second phase, a more suitable index, called the average sparsity concentration index (ASCI), is proposed, which dramatically promotes the clustering accuracy compared to the conventionally utilized SCI index. In addition, the subspace evolution detection model based on the Page-Hinkley test is proposed where the appearing, disappearing, and recurring subspaces can be detected and adapted. Extinct experiments on real-world data streams show that the EDSSC outperforms the state-of-the-art online SC approaches.

Low Rank Representation Subspace Clustering Algorithm Based on Hessian Regularization and Non Negative Constraints

“…LRR is built on the premise of union sampling on multiple low rank linear subsets, and LRR also has the property of self-expression, that is, each sample is expressed in its own form. Unlike SSC, which selects the least number of features from multiple expressions, LRR aims to find the minimum rank of all data [14][15]. Therefore, compared with SSC, LRR can better control the overall structure of sample points, that is, by applying low rank constraints to the expression matrix, it aggregates data points with high correlation together, thereby achieving overall clustering.…”

Two-Stage Sparse Representation Clustering for Dynamic Data Streams

“…Moreover, these techniques are naturally adapted to the evolving environment of data streams. In addition, EDSSC introduces a subspace structure evolution detection model to detect the appearing, disappearing, and recurring subspaces [22]. In particular, it uses a singular-based Laplacian matrix decomposition to automatically estimate the number of subspaces.…”

“…However, it is insufficient to characterize the intrinsic structures of high-dimensional data objects in data streams. This motivates the development of alternative techniques for exploiting the intrinsic characteristics of the highdimensional data objects [10], [22], [23], [23], [24]. For example, Krleža1 et al proposed a statistical hierarchical clustering (SHC) algorithm that uses statistical inference based on statistical distances to estimate statistical distributions on data streams [10].…”

“…For example, Krleža1 et al proposed a statistical hierarchical clustering (SHC) algorithm that uses statistical inference based on statistical distances to estimate statistical distributions on data streams [10]. Sui et al presented an evolutionary dynamic sparse subspace clustering (EDSSC) algorithm based on sparse representation to address the problem of concept drift [22]. Li et al proposed an efficient self-adaptive online data stream (ESA-Stream) clustering algorithm that self-adaptively learns parameters to cluster data streams in a dynamic online environment [25].…”

Research on concept drift data stream classification based on adaptive online broad learning system

“…Data stream mining is widely used in various fields such as weather forecasting, power prediction, stock trading, and network intrusion detection. In recent years, research on data streams has received considerable attention with the aim of improving the generalization performance of online learning models to adapt to the real-time distribution of data flow [1]. Since the data distribution of data streams may change over time, that is, there is concept drift, the model needs to be continuously updated to adapt to the new data stream environment [2].…”