An Experiment with Fuzzy Sets in Data Mining

Authors:
David L. Olson;Helen Moshkovich;Alexander Mechitov
Affiliations:
University of Nebraska, Department of Management, Lincoln, NE 68588-0491, USA;Montevallo University, Comer Hall, Montevallo, AL 35115, USA;University of Nebraska, Department of Management, Lincoln, NE 68588-0491, USA
Venue:
ICCS '07 Proceedings of the 7th international conference on Computational Science, Part II
Year:
2007

Citing 4
Cited 0

Probabilistic reasoning in intelligent systems: networks of plausible inference

Probabilistic reasoning in intelligent systems: networks of plausible inference
Classification with Degree of Membership: A Fuzzy Approach

ICDM '01 Proceedings of the 2001 IEEE International Conference on Data Mining
Mining Fuzzy Rules in A Donor Database for Direct Marketing by a Charitable Organization

ICCI '02 Proceedings of the 1st IEEE International Conference on Cognitive Informatics
A systematic approach to the assessment of fuzzy association rules

Data Mining and Knowledge Discovery

Quantified Score

Hi-index	0.00

Visualization

Abstract

Fuzzy modeling provides a very useful tool to deal with human vagueness in describing scales of value. This study examines the relative error in decision tree models applied to a real set of credit card data used in the literature, comparing crisp models with fuzzy decision trees as applied by See5, and as obtained by categorization of data. The impact of ordinal data is also tested. Modifying continuous data was expected to degrade model accuracy, but was expected to be more robust with respect to human understanding. The degree of accuracy lost by See5 fuzzification was minimal (in fact more accurate in terms of total error), although bad error was worse. Categorization of data yielded greater inaccuracy. However, both treatments are still useful if they better reflect human understanding. An additional conclusion is that when categorizing data, care should be taken in setting categorical limits.