Evaluation of Adaptive Genetic Algorithm to Environmental Modeling of Peponapis and Cucurbita

Abstract

The environmental modeling could be performed combining environmental data and species occurrence points in order to create distribution models, and this requires non trivial algorithms. The objective of this work is comparing GARP, one of the most used algorithms, with ADAPTGARP, that implements adaptive tools. The experiments considered some species of Peponapis and Cucurbita, respectively, pollinator insects and their plants. The maps with potential species distribution resulting from comparative experiments are very similar. The variations of AUC values, an accepted measure of accuracy on modeling, are also within the tolerance limits. The experiment shows that the algorithms are equivalent and the adaptive tools are adequate. Future works could be developed using adaptivity, such as including generic characteristics on the genetic algorithm and new types of treatment, looking for more accurate models.