Fuzzy Clustering

Classical clustering techniques are well known. They allow the classification of a number of observations into a few groups on the basis of their characteristics.

 

The classifications that result have a number of weaknesses:

• They are black and white

• They often use restricted amounts of information

• It can be hard to produce a sensible classification

Volterra has carried out a lot of empirical work applying clustering techniques using fuzzy logic. These include:

• Allow overlap between different segments

• Can handle more useful information

• Generally produce fewer segments

• Provide a sense of the strength of the difference between groups

• Can analyse how much change is necessary for an individual observation to move to a different category

The techniques are relevant to a number of commercial applications and have also been used to analyse university performance for the Times Higher Educational Supplement. Work using fuzzy clustering includes: Economic Geography of Great Britain, Retail Locations, Universities, Social & Economic Structure of the South East.