ParRescue: Scalable Parallel Algorithm and Implementation for Biclustering over Large Distributed Datasets

Biclustering refers to simultaneously capturing correlations present among subsets of attributes (columns) and records (rows). It is widely used in data mining applications including biological data analysis, financial forecasting, and text mining. Biclustering algorithms are significantly more complex compared to the classical one dimensional clustering techniques, particularly those requiring multiple computing platforms for large and distributed data sets. In this paper, we develop an efficient scalable algorithm, referred to as ParRescue(Parallel Residue Co-clustering), that is capable of performing biclustering on extremely large or geographically distributed data sets. ParRescue divides the cluster tasks among processors with minimal communication costs thus making it scalable over large number of computing nodes. The proposed implementation is based on an existing sequential approach that has been modified for amenable parallel implementation. The proposed ParRescue algorit...

• Type: Conference
• Year: 2006
• Where: ICDCS
• Authors: Jianhong Zhou, Ashfaq A. Khokhar
• Keywords: Data Mining Applications, Distributed And Parallel Computing, Distributed Data Sets, Efficient Scalable Algorithm, ICDCS 2006