INLEN: a methodology and integrated system for knowledge discovery in databases
Kenneth A. Kaufman · George Mason University eBooks · 1998
This thesis presents a methodology for multistrategy knowledge discovery from databases, and its experimental validation through the implementation and testing of the INLEN system. The presented methodology is based on the integration of diverse machine learning operators with traditional data analysis tools into a multi-operator environment. Among the issues that must be confronted in order to implement such an architecture are enabling the machine learning tools to handle the structured or numerical attribute domains they are likely to encounter, developing a sufficiently rich knowledge representation system to allow the different operators to pass necessary information to each other and to human data analysts, and the creation of a means by which a system implemented under such an architecture can function semi-autonomously, rather than requiring a domain expert to analyze the output from each step in order to determine how next to proceed. The methodology outlined here is implemented in a system called INLEN, based on the AURORA system developed by International Intelligent Systems, Inc. INLEN integrates a database, a knowledge base, and machine learning methods within a uniform user-oriented framework. A variety of domain-independent machine learning programs are incorporated into the system to serve as high-level knowledge generation operators. These operators can generate diverse kinds of knowledge about the properties and regularities existing in the data. Examples of INLEN's abilities and performance are presented. The goals and contributions of this research are to investigate the problems presented to machine learning methods by data exploration problems, to enhance the exploitation of background and discovered knowledge by machine learning programs, to develop control systems for integrating knowledge discovery technologies, and to utilize these advances by implementing a system capable of contributing knowledge and understanding in complex real-world domains.