In many disciplines cluster analysis is pattern detection rather than pattern recognition. Pattern detection is used to try to answer questions like "Are there groups of cases that are similar within the group, and the groups are distinct from each other?" Pattern recognition is used to try to answer questions like "How are these groups of cases distinct from each other?" (e.g., discriminant function analysis) and "How well to these cases fit into these groups?". Frequently clustering is done as a first step before classifying other cases.
I have used clustering in studying types of schizophrenics, types of classroom environments, types of students, types of US counties, and data mining. A scientific organization that deals with methods like clustering, discrimination, multidimensional scaling, mixed scaling and clustering, etc. is the Classification Society of North America. This is a group where a great deal to disciplinary miscegenation is committed. Here psychologists, biologists, information retrieval people, statisticians, sociologists, computer scientists, astronomers, market researchers, cartographers, etc. share methods and results in approaching these kinds of questions. at http://www.pitt.edu/~csna/index.html you'll see The Journal of Classification is ranked 10th highest in impact of all journals in mathematics for the period 1981-2000! The Classification Society of North America Annual Meeting (CSNA 2003) will be June 12-15, 2003 at the Doubletree Hotel, Tallahassee, Florida. various links to see what has been covered in the Journal and in meetings. Hope this helps. Art [EMAIL PROTECTED] Social Research Consultants University Park, MD USA (301) 864-5570 Martin wrote: > Cluster analysis is a statistical technique that is used in many > different domains. > > Image segmentation, text categorization and micro-array data analysis > all involve cluster analysis, and color clustering in a colored image > is also important. > > Can you suggest some other important applications of cluster analysis > outside the domain of pattern recognition? Pointers in the > literature are very much appreciated. > > Thank you!! . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
