| |
Procedures for combining individual observations into categories. Classification involves splitting a population into mutually exclusive categories on predetermined criteria, either deductively (using a previously determined set of classes) or inductively (finding the best set of classes for the particular data set). Some procedures take the entire population and divide it into classes whereas others begin with the individuals and group them into classes: in both cases, the outcome is a hierarchy of classes (e.g. all people divided by gender; each gender group divided by ethnic origin; each gender and ethnic group divided by age; etc.) With inductive classifications, the guideline is usually that each member of a class should be more like all of the other members of that class than it is like the members of any other: the classes are thus internally homogeneous and externally heterogeneous.
Regionalization (see region and regional geography) is a special case of the more general procedure of classification. The individuals comprising the population to be classified are areas and the resulting classes (regions) must form contiguous spatial units. Because of this additional criterion, regions defined for a population of areas may not be as internally homogeneous as a classification of the same areas without the contiguity constraint.
A range of computer algorithms has been devised to produce classifications and regionalizations. Most have precise inter-class and inter-region boundaries, but Openshaw and Openshaw (1997) have suggested the use of fuzzy logic to indicate the probability that an individual is a member of any particular class. (RJJ)
References and Suggested Reading Johnston, R.J. 1976: Classification in geography. Concepts and techniques in modern geography 6. Norwich: Geo-Books. Openshaw, S. and Openshaw, C. 1997: Artificial intelligence in geography. Chichester and New York: John Wiley. |
|
