DBSC: A Dependency-Based Subspace Clustering Algorithm for High Dimensional Numerical Datasets

Abstract. We present a novel algorithm called DBSC, which finds subspace clusters in numerical datasets based on the concept of ”dependency”. This algorithm employs a depth-first search strategy to find out the maximum subspaces. Next the clusters are mined within those maximum subspaces. Our algorithm shows great scalability and high efficiency for high-dimensional datasets, it is also robust to outliers and requires no pre-conception of the dataset. We are providing a conjunction representative for each cluster. The experiment results both on synthetic and real datasets show this algorithm is very effective and promising.

• Type: Conference
• Year: 2007
• Where: AUSAI
• Authors: Xufei Wang, Chunping Li
• Keywords: Algorithm, Algorithm Called Dbsc, Artificial Intelligence, AUSAI 2007, Maximum Subspaces