Clusan: A Knowledge Base for Approximate Reasoning in Exploratory Data Analysis

• Published in: Machine Intelligence and Pattern Recognition Vol. 7; pp. 395 - 411
• Main Author: Backer, E.
• Format: Book Chapter
• Language: English
• Published: 1988
• ISBN: 9780444871374; 0444871373
• ISSN: 0923-0459
• DOI: 10.1016/B978-0-444-87137-4.50033-3

Cluster analysis as a scientific tool to unravel data is characterized by multiple statistical testing, validation and complex reasoning. Today it is felt natural to associate such a reasoning process directly with expert systems. This paper is a result of an attempt to develop a knowledge base (CLUSAN1) for the expert system shell DELFI2 to facilitate the user to obtain validated results of an explorative data analysis. As a result, the expert system approach appears to be particularly suitable for potential users who are non-experts but familiarized with the subject matter. In this paper we identify the problem, briefly discuss the requirements and the architecture of the system as well as the representation of knowledge and the use of external routines. More in detail, we discuss a problem example in the sub-context “clustering tendency” based on specific knowledge of k-means and MST-algorithms.