Autoria: LAZZARETTI, A. T.; SCHNEIDER, V. R.; WIEST, R.; LAU, D.; FERNANDES, J. M. C.; FRAISSE, C. W.; CERBARO, V. A.; KARREI, M. Z.

Resumo: Resumo: Os insetos ao atingirem um determinado nível populacional podem causar danos às plantas, sendo considerados pragas. Afídeos ou pulgões apresentam um alto potencial biótico e podem causar diferentes tipos de dano às plantas. Fatores meteorológicos como precipitações, ventos e temperatura interferem no crescimento populacional destes insetos. Este trabalho aplicou diferentes técnicas de machine learning com o objetivo de verificar a correlação existente entre variáveis meteorológicas e a dinâmica populacional dos afídeos. Foram implementados 4 (quatro) modelos obtendo-se as acurácias de 11,4% para Regressão Linear; 26,4% para o modelo de Rede Neural Artificial; 29,3% para Árvore de decisão e 41,4% para random forest. Abstract: Insects have an important degree of collaboration for the maintenance of the ecosystem on the planet. However, after reaching a certain population level and causing damage to plants, some insects are considered as pests and represent a threat to agriculture. Aphids insects that has characteristics to reach this state as it has a high biotic potential and can cause different types of damage to plants. Climatic data as precipitation, winds and temperatures affect the population quantity of these insects. Therefore, this work proposes to apply different machine learning techniques with the objective to verify the existing correlation between climatic variables and the population dynamics of aphids. It can be concluded that variables such as precipitation, temperature, number of days when it rains in the week and climatic phenomena such as El niño and La niña have an influence on the aphid population. During the work, four models were developed in order to predict the population of these insects. The accuracy of the prediction model developed were 11.4% for Linear Regression; 26.4% for the Artificial Neural Network model; 29.3% for Decision Tree and 41.4% for Random Forest.

Ano de publicação: 2023

Tipo de publicação: Artigo de periódico

Unidade: Embrapa Trigo

Palavras-chave: Afídeo, Análise de Dados, Artificial neural networks, Decision tree, Exploratory Data, Extração de conhecimento, Knowledge extraction, Linear Regression, Plant pests, Plants (botany), Praga de Planta, Pulgão, Random Forest, Redes Neurais Artificiais, Regressão Linear, Árvore de decisão