Relational Case-based Reasoning for Carcinogenic Activity Prediction

摘要

Lazy learning methods are based on retrieving a set of precedent cases similar to a new case. An important issue of these methods is how to estimate the similarity among a new case and the precedents. Usually, similarity measures require that cases have a prepositional representation. In this paper we present Shaud, a similarity measure useful to estimate the similarity among relational cases represented using featureterms. We also present results of the application of Shaud forsolving classification tasks. Specifically we used Shaud for assessingthe carcinogenic activity of chemical compounds in the Toxicology dataset.