Efficient mining of distance‐based subspace clusters

Abstract

Traditional similarity measurements often become meaningless when dimensions of datasets increase. Subspace clustering has been proposed to find clusters embedded in subspaces of high dimensional datasets.Many existing algorithms use a grid based approach to partition the data space into nonoverlapping rectangle cells, and then identify connected dense cells as clusters. The rigid boundaries of the grid based approach may cause a real cluster to be divided into several small clusters. In this paper, we propose to use a sliding window approach to partition the dimensions to preserve significant clusters. We call this model nCluster model. The sliding window approach generates more bins than the gridbased approach, thus it incurs higher mining cost. We develop a deterministic algorithm, called MaxnCluster, to mine nClusters efficiently. MaxnCluster uses several techniques to speed up the mining, and it produces only maximal nClusters to reduce result size. Non-maximal nClusters are pruned without the need of storing the discovered nClusters in the memory, which is key to the efficiency of MaxnCluster. Our experiment results show that MaxnCluster can produce maximal nClusters efficiently and accurately.

A Critical Review of the Applications and AI Techniques for Anomaly Detection

“…The subspaces of elevated information are often where clusters are found. Liu, Sim, Li and Wong [14] came to the conclusion that it is possible to recommend more relevant clusters that are unique to a certain subspace by examining the nature of data. These data sets have missing or low-dimensional subspaces, making it difficult to detect abnormalities.…”

The blind men and the elephant: on meeting the problem of multiple truths in data from clustering and pattern mining perspectives

“…2 The pioneer approach to finding all clusters in all subspaces coining the term 'subspace clustering' for this specific task has been CLIQUE (Agrawal et al 1998). Numerous variants have been proposed, e.g., by Cheng et al (1999), Nagesh et al (2001), Kailing et al (2004a), Assent et al (2007Assent et al ( , 2008, Moise and Sander (2008), Müller et al (2009a), Liu et al (2009). Like for frequent itemset mining, one can question the original problem formulation of finding 'all clusters in all subspaces', as by retrieving a huge and highly redundant set of clusters the result will not be very useful or insightful.…”

GPUMAFIA: Efficient Subspace Clustering with MAFIA on GPUs

“…MAFIA [3] was introduced shortly thereafter. Alternative subspace clustering algorithms are: SeqClus [7], LCM-nCluster [8], Maxncluster [9], and DiSH (Detecting Subspace cluster Hierarchies) [10]. Alternative subspace clustering algorithms are: SeqClus [7], LCM-nCluster [8], Maxncluster [9], and DiSH (Detecting Subspace cluster Hierarchies) [10].…”

“…A good survey of existing subspace clustering algorithms can be found in [2] which, however, does not include a comparison of performance or quality. In [9] Maxncluster is compared to MAFIA and STATPC. For example, in [11] MAFIA and FINDIT are compared.…”