Applying statistical knowledge to database analysis and knowledge base construction

A software package, called KnowledgeSeeker, that combines statistical problem-solving algorithms with an interactive, graphically-oriented decision-tree display is discussed. It has a browser to enable end users to rapidly extract high-quality decision-making information from a database in the form of a decision tree or as a knowledge base of rules. KnowledgeSeeker captures statistical problem-solving knowledge in several important respects and embeds this knowledge in software algorithms that can be run in interactive or automatic mode. This allows end users to quickly build high-quality, accurate, and reliable decision trees and to readily explore a wide range of potentially significant database effects through a preview mechanism. The graphic decision-tree display takes advantage of statistical grouping operations to produce a compact, readily comprehensible summary of the database field relationships and resulting partitions of the decision tree. Some applications of the program are described.<
>