Confirmation Rule Sets

The concept of confirmation rule sets represents a framework for reliable decision making that combines two principles that are effective for increasing the predictive accuracy: consensus in an ensemble of classifiers and indecisive or probabilistic predictions in cases when reliable decisions are not possible. The confirmation rules concept uses a separate classifier set for every class of the domain. In this decision model different rules can be incorporated: either those obtained by applying one or more inductive learning algorithms or even rules representing human encoded expert domain knowledge. The only conditions for the inclusion of a rule into the confirmation rule set are its high predictive value and relative independence of other rules in the confirmation rule set. This paper introduces the concept of confirmation rule sets, together with an algorithm for selecting relatively independent rules from a set of all acceptable confirmation rules and an algorithm for the systematic construction of a set of confirmation rules.

[1]  Nada Lavrac,et al.  The Multi-Purpose Incremental Learning System AQ15 and Its Testing Application to Three Medical Domains , 1986, AAAI.

[2]  Ronald L. Rivest,et al.  Learning complicated concepts reliably and usefully , 1988, Annual Conference Computational Learning Theory.

[3]  P. Boland.  Majority Systems and the Condorcet Jury Theorem , 1989 .

[4]  Michael J. Pazzani,et al.  Trading off coverage for accuracy in forecasts: Applications to clinical data analysis , 1994 .

[5]  Yoav Freund,et al.  Experiments with a New Boosting Algorithm , 1996, ICML.

[6]  J. Ross Quinlan,et al.  Bagging, Boosting, and C4.5 , 1996, AAAI/IAAI, Vol. 1.

[7]  Kagan Tumer,et al.  Error Correlation and Error Reduction in Ensemble Classifiers , 1996, Connect. Sci..

[8]  Heikki Mannila,et al.  Fast Discovery of Association Rules , 1996, Advances in Knowledge Discovery and Data Mining.

[9]  I. Kononenko,et al.  Machine learning improves the accuracy of coronary artery disease diagnostic methods , 1997, Computers in Cardiology 1997.

[10]  Wynne Hsu,et al.  Integrating Classification and Association Rule Mining , 1998, KDD.

[11]  Jiri Matas,et al.  On Combining Classifiers , 1998, IEEE Trans. Pattern Anal. Mach. Intell..

[12]  Nada Lavrac,et al.  A Relevancy Filter for Constructive Induction , 1998, IEEE Intell. Syst..

[13]  Nada Lavrac,et al.  Experiments with Noise Filtering in a Medical Domain , 1999, ICML.

[14]  Igor Kononenko,et al.  Analysing and improving the diagnosis of ischaemic heart disease with machine learning , 1999, Artif. Intell. Medicine.

[15]  Nada Lavrac,et al.  Diagnostic Rules of Increased Reliability for Critical Medical Applications , 1999, AIMDM.