Incremental learning in attributenets with dynamic reduct and IQuickReduct

Abstract

Incremental learning is becoming more essential in the real world problems in which a decision system is being updated frequently. AttributeNets is a classifier whose representation allows updating the classifier when new data is added incrementally. In this paper the impact of reduct on the performance of AttributeNets as an Incremental Classifier is investigated. This philosophy has been demonstrated by adopting two varieties of reducts, namely dynamic reduct and IQuickReduct. These reducts were used to study the capability of AttributeNets for classification with reduced attributes. © 2011 Springer-Verlag.