Avaliação de patogenicidade do fungo entomopatogênico Beauveria bassiana em formigas do gênero Atta spp. utilizando aprendizado de máquina

Abstract

The application of pesticides increases agricultural productivity, but frequent and intensive use generates multiple unfavorable externalities. Biological Control, emerged as an alternative to chemical methods, and currently shows significant numbers in the world agricultural market. This method uses natural enemies, in order to minimize the presence of harmful residues to human health, fauna and flora species and the environment. The entomopathogenic fungus Beauveria bassiana is commonly used to control insect pests, however, a recommended dosage to control leaf-cutting ants of the genus Atta spp., has not yet been reported. In view of this, this work evaluated the pathogenicity of this fungus through the use of Machine Learning. The data presented were extracted from an Artificial Biological Control work, carried out under laboratory conditions in the year 2018. The bioassays were carried out with the application of the biopesticide Boveril®. As there is no recommended dosage of Beauveria for Atta spp., the concentrations were prepared according to the recommendation for the coffee berry borer (Hypothenemus hampei). From the results obtained, a dataset was built, bringing information about the dosage used in each treatment, and how many days it took to exterminate a certain percentage of ants. The decision tree was generated by the Weka software, version 3.6.11. The values obtained through Machine Learning, found a Boveril® dosage pattern and a significant result that was able to combine efficiency in the death of the ants in as many days as possible.