Pattern recognition system based on decision trees and fuzzy logic: Anti-HIV molecules application

Several works structure activity relationship (SAR) of anti-HIV molecules (Human Immunodeficiency Virus) were studied by different statistical methods and non-linear models (neural networks). But few studies have used the heuristic methods. In this work, we are interested to study this relationship by fuzzy logic and decision trees. The resulting model explain SAR with only tow rules described by three of 7 molecular descriptors. This rules generalize the 79 compounds studied. Decision trees show good performance in the learning and prediction phases.