The spatial auto-regression model (SAM) is a popular spatial data mining technique which has been used in many applications with geo-spatial datasets. However, serial procedures for estimating SAM parameters are computationally expensive due to the need to compute all the eigenvalues of a very large matrix. We propose a parallel formulation of the SAM parameter estimation procedure using data parallelism and hybrid programming technique. Experimental results on an IBM Regatta show that the proposed parallel formulation achieves a speedup of up to 7 on 8 processors. We are developing algebraic cost models to analyze the experimental results to further improve the speedups.

Keywords: Spatial Auto-regression, Spatial Auto-correlation, Parallel Formulation, Spatial Data Mining

DATE: June 23, 2004, Wednesday @ 15:40
PLACE: EA-409