A hybrid seagull optimization algorithm for chemical descriptors classification

Cheminformatics is a field of research that explores the correlation between chemical compound data sets and the metrics used in drug design to evaluate the similarity. Many classical-methods are applie-d in prediction the drug design that aren't efficient and not effective. This article introduces a classification approach called SOA-SVM that is an hybrid version of the Seagull optimization Algorithm (SOA) combined with Support Vector Machines (SVM) and designed to select the descriptors for chemical compound tasks. The operators used in the exploration and exploitation phases in SOA are modified in order to identify the desired features that permits to enhance the accuracy. For experimental purposes they are employed two data sets, namely, MonoAmine Oxidase (MAO) and QSAR Biodegradation. The results helps to venfy that good performance of the proposal, that is able to find the best solutions to the feature selection problem with a high accuracy in comparison with other methodologies.