Lazy Induction of Descriptions Using Two Fuzzy Versions of the Rand Index

In this paper we introduce an extension of the lazy learning method called Lazy Induction of Descriptions (LID). This new version is able to deal with fuzzy cases, i.e., cases described by attributes taking continuous values represented as fuzzy sets. LID classifies new cases based on the relevance of the attributes describing them. This relevance is assessed using a distance measure that compares the correct partition (i.e., the correct classification of cases) with the partitions induced by each one of the attributes. The fuzzy version of LID introduced in this paper uses two fuzzy versions of the Rand index to compare fuzzy partitions: one proposed by Campello and another proposed by Hüllermeier and Rifqi. We experimented with both indexes on data sets from the UCI machine learning repository.