Discovery of Predictive Neighborly Rules from Neighborhood Systems

Document Type

Contribution to Book

Publication Date

7-1-2013

Publication Title

Proceedings of the International Conference on Information and Knowledge Engineering (IKE)

Abstract

Georgia Southern University faculty member Ray R. Hashemi authored "Discovery of Predictive Neighborly Rules from Neighborhood Systems" in International Conference on Information and Knowledge Engineering (IKE'13).

The use of "data closeness" for clustering, concept generalization, and imprecise query processing has been frequently reported in the literature. In this article, however, the authors have introduced the use of "data closeness" for building a prediction tool. To do so, they: (1) Generate the workable neighborhood system for every record, Ri, of a training set, (2) build and expand the "record tree" for Ri, using its workable neighborhood system, (3) Extract a neighborly rule from each expanded record tree, and (4) Use the rules for prediction. The empirical results revealed that, the predictive power of the neighborly rules is comparable with that of ID3 and Rough Sets.

Share

COinS