Spatial Data Mining:
Consolidation and Renewed Bearing
April 22, 2006

call for papers
challenge: pandemic preparedness

News: the agenda for the workshop, including the list of accepted papers, is now available.

Spatial data sets are at the heart of a variety of scientific and engineering domains, from astrophysics to computational fluid dynamics to robotics. Rapid advances in simulation and experimentation in these domains are yielding an increasing reliance on efficient and effective spatial reasoning algorithms. New applications in other domains, such as aerodynamics, scientific computing, RFID, sensor and actuator networks, and structural bioinformatics, are additionally being cast in terms of mining and reasoning about spatial data. These developments demand effective cross-fertilization and consolidation of computational techniques from qualitative reasoning, data mining, scientific computing, and statistical methodology, in the context of significant applications.

wireless portrait pocket

Significant spatial data mining research is being conducted in different communities driven by complementary goals and pursuing complementary approaches. The confluence of research agendas is exhibited in several trends:

  1. Researchers working with spatial datasets are now increasingly turning toward model-based analyses for making more effective use of data. Often this takes the form of incorporating domain knowledge to more effectively cull information from a given dataset. This trend is evident, for example, in feature-based extraction of weather patterns, qualitative reasoning about scientific computations, and interpretation of high-throughput bioinformatics experiments.
  2. A significant body of theoretical work has emerged in the statistical modeling/machine learning literature under the broad title of `Gaussian Processes.' This encompasses new probabilistic models for regression and classification, reformulations of classical machine learning algorithms, design of experiments theory, and surrogate models. Interestingly, these ideas have their roots in spatial statistics and, needless to say, can naturally be targeted at studying spatial phenomena.
  3. Important methodological issues must be addressed in the areas of integrating data collection and mining (`closing-the-loop'), and reasoning when data is scarce. Nowhere is this aspect more important than in spatial domains, which enjoy a range of data acquisition and instrumentation tools, and where data collection can be costly.

The purpose of the SIAM-DM 2006 Workshop on Spatial Data Mining is to bring together researchers from areas of spatial data mining, spatial modeling, and applications, and provide a forum for exchanging ideas, fostering collaborations, and gaining momentum.