The cotton boll weevil Anthonomus grandis grandis Boh. (Coleoptera: Curculionidae) is a destructive pest of cotton and can cause losses of up to 100%. Understanding and predicting the seasonal dynamics of pests makes it possible to plan sampling and apply control methods more efficiently. Artificial neural networks (ANNs) are machine learning tools with high predictive power. Therefore, this work aimed to determine a prediction model for the seasonal dynamics of A. grandis grandis in cotton crops using ANN. We used pest density data obtained in fields located in Brazil that were collected for five years. 1716 ANN were determined. The selected ANN used meteorological data from 20 days before evaluating the pest density, six neurons in the hidden layer, logistic activation function and Rprop learning algorithm. ANN predictions showed a high correlation (Pearson correlation = 0.865) with crop pest densities. Among the ANN predictors, the duration of the reproductive stage, air temperature and rainfall had positive effects on pest density. The opposite occurred with relative air humidity. Thus, the ANN determined in this work is promising to be used to predict the seasonal dynamics of A. grandis grandis.
Oliveira et al. (Thu,) studied this question.