Some Late-Breaking News from the Data Mines and a Preview of the KOALA System: A Prospector's Report

It has been widely advertised that the numerous large databases which exist in the various industries, administrative offices and in the public domain (e.g. the world-wide-web) would indeed be very valuable data mines from which important and previously unknown knowledge could be harvested when machine learning procedures would be applied as mining tools. The present research evaluates the prospects of discovering such knowledge from industrial databases. Three different databases are considered and three different machine learning tools (conceptual clustering, neural net, and inductive logic programming) are applied in an experimental fashion. From these experiences it could be concluded that the tool box philosophy has severe limitations in highly structured industrial application areas. It was thus suggested that higher order conceptualizations of machine learning should be developed which are easier to apply and understand by the user. A preview of the KOALA system which is currently under development is then presented. By applying constraint satisfaction over hierarchically structured domains, the KOALA system allows the user to have his own machine learning application being configured according to the domain ontologies and the specific needs of a given field of application.

[1]  Franz Schmalhofer,et al.  Unified approach to learning for complex real-world domains , 1995, Appl. Artif. Intell..

[2]  William Frawley,et al.  Knowledge Discovery in Databases , 1991 .

[3]  Derek H. Sleeman.  Towards a Technology and a Science of Machine Learning , 1994, AI Commun..

[4]  Luc Steels,et al.  A Future for Knowledge Acquisition , 1994, Lecture Notes in Computer Science.

[5]  Ralph Bergmann,et al.  Learning program abstractions: Formal model and empirical validation , 1993 .

[6]  John Skilling,et al.  Maximum Entropy and Bayesian Methods , 1989 .

[7]  Andreas Abecker,et al.  An Environment for Exploring and Validating Declarative Knowledge , 1995, LPE.

[8]  Franz Schmalhofer,et al.  Cooperative knowledge evolution for complex domains , 1995 .

[9]  L. D. Raedt.  Interactive theory revision: an inductive logic programming approach , 1992 .

[10]  Y. Kodratoff,et al.  From machine learning towards knowledge discovery in databases , 1995 .

[11]  Knut Hinkelmann,et al.  Knowledge-Base Evolution for Product and Production Planning , 1994, AI Commun..

[12]  Brian R. Gaines,et al.  Machine learning and knowledge acquisition: integrated approaches , 1995 .

[13]  Tom Khabaza,et al.  Data mining with Clementine , 1995 .

[14]  F. Schmalhofer,et al.  Chapter 10 Learning Program Abstractions: Model and Empirical Validation , 1993 .

[15]  Gerhard Strube,et al.  The Cognitive Psychology of Knowledge , 1993 .

[16]  J. Stutz,et al.  Autoclass — A Bayesian Approach to Classification , 1996 .