Geographia Polonica (1989) vol. 57

Cluster analysis and large data sets: a case study of farming systems in France

John William Aitchison

Geographia Polonica (1989) vol. 57, pp. 13-22 | Full text

With the increasing availability of large, spatially-indexed data banks and the emergence of sophisticated geo-processing systems, agricultural geographers are now in a position to undertake much more detailed and wide-ranging investigations into the typological and regional structure of farming systems. These technological develop-ments also allow a more experimental and critical stance to be adopted in studies of a taxonomic (classificatory) nature. This is important since classification is essentially an exploratory process — a search for meaningful or revealing patterns of order within complex multivariate data sets. It is not a search for single solutions that can be regarded as "definitive" or "true". Typologies and regionalizations can be effected in many different ways, and it behoves would be taxonomists to test and evaluate a range of classificatory models, and to justify the categorizations that are eventually selected for subsequent interpretation. Needless to say, the fact that it is now a relatively simple matter to generate maps and plots of classified units aids this process of experimen-tation considerably, for from a geographer's point of view it is often the meaningfulness of the resultant spatial distributions that is of paramount diagnostic importance. It is not possible here to examine these various issues in great detail; the more limited aim is to consider the general problem of classifying large sets of agricultural data. In so doing, particular emphasis will be placed on "iterative partitioning" methods of cluster analysis.

John William Aitchison, Department of Geography, University College of Wales, Aberystwyth, UK