Modeling uncertainty in knowledge discovery for classifying geographic entities with fuzzy boundaries

Abstract

Boosting is a machine learning strategy originally designed to increase classification accuracies of classifiers through inductive learning. This paper argues that this strategy of learning and inference actually corresponds to a cognitive model that explains the uncertainty associated with class assignments for classifying geographic entities with fuzzy boundaries. This paper presents a study that adopts the boosting strategy in knowledge discovery, which allows for the modeling and mapping of such uncertainty when the discovered knowledge is used for classification. A case study of knowledge discovery for soil classification proves the effectiveness of this approach. © 2006 Springer-Verlag Berlin Heidelberg.