A version of the NGE model suitable for fuzzy domains
Journal of Intelligent & Fuzzy Systems: Applications in Engineering and Technology
| Hi-index | 0.00 |
In the original proposal of the NGE (NestedGeneralized Exemplar) system, the induction of a conceptis based on an initial set of training examples (namedseeds) that are randomly chosen. The number of examplesin this set is arbitrary, generally determined by the userof the system. It can be seen empirically, that the finalresults are influenced by the initial choice of the seeds.The work described in this paper proposes andinvestigates other alternative methods for choosing seedsand empirically evaluates their impact on the learningresults in seven knowledge domains, as far as accuracyand number of expressions describing the concepts areconcerned. In spite of the additional time investmentassociated with using a clustering method and, assumingthat accuracy of the induced concept is of majorimportance, experiments have shown that choosing theinitial set of seeds as the center of clusters can be the bestoption.