Scalable Parallel Clustering for Data Mining on Multicomputers

Abstract

Abstract. This paper describes the design and implementation on MIMD parallel machines of P-AutoClass, a parallel version of the AutoClass system based upon the Bayesian method for determining optimal classes in large datasets. The P-AutoClass implementation divides the clustering task among the processors of a multicomputer so that they work on their own partition and exchange their intermediate results. The system architecture, its implementation and experimental performance results on different processor numbers and dataset sizes are presented and discussed. In particular, efficiency and scalability of P-AutoClass versus the sequential AutoClass system are evaluated and compared.

Clustered Distributed Index for Efficient Text Retrieval Using Threads

“…Surprisingly our results show that by using the concept of clustering and parallelism, search becomes more cost effective, time effective and the quality of the search becomes accurate. Our results show that this strategy is able to cause efficient performance both in large scale and small scale search engines [16].…”

“…developed scalable parallel clustering models for data mining on multi-computers in their research paper in [16]. They designed & implemented on MIMD parallel machines of PAutoClass, a parallel version of the AutoClass system based upon the Bayesian method for determining optimal classes in large datasets.…”

Parallelism in Knowledge Discovery Techniques

“…An example of parallel implementation of a clustering algorithm is P-CLUSTER [10]. Other parallel clustering algorithms are discussed in [5], [12], and [7]. In particular, in [7] an SPDM implementation of the AutoClass algorithm, named P-AutoClass is described.…”

“…Other parallel clustering algorithms are discussed in [5], [12], and [7]. In particular, in [7] an SPDM implementation of the AutoClass algorithm, named P-AutoClass is described. The paper shows interesting performance results on distributed memory MIMD machines.…”

A Local Facility Location Algorithm for Sensor Networks

“…Nevertheless, it is easy to see how distributed formulation of related hill-climbing algorithms such as k-means and k-median clustering [5,7,6] can be adapted to solve distributed FLP. We note, however, that all previous work on distributed clustering assumes tight cooperation and synchronization between the processors containing the data, and a central processor that collects the sufficient statistics needed in each step of the hill-climbing heuristic.…”