Analysis of pest incidence on apple trees validated by unsupervised machine learning algorithms

• Autor(a) principal: Speranza, Eduardo Antonio
• Data de Publicação: 2022
• Outros Autores: Grego, Célia Regina, Gebler, Luciano
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Engenharia na Agricultura
• Texto Completo: https://periodicos.ufv.br/reveng/article/view/12919
• topic: Pest management; Geostatistics; Unsupervised Machine Learning; Orchards; Apples; Pest management; Unsupervised Machine Learning; Orchards; Apples

Integrated pest control is a practice commonly used in apple orchards in southern Brazil. This type of management is an important tool to help improve quality and increase yields. This study aimed to identify areas with higher and lower incidence of aerial pests in a commercial apple orchard, regarding data collected from three different crops using georeferenced traps. Geostatistical analyses were performed, based on the modeling of semivariograms and spatial interpolation using the kriging method; and clustering, based on specific unsupervised machine learning algorithms for count data. The algorithms were selected from measures of stability, connectivity and homogeneity, seeking to identify areas with different incidence of pests that could help farmer decision making regarding insect population control using pesticides. The geostatistical analysis verified the presence of individual pest infestations in specific sites of the study area. Additionally, the analysis using machine learning allowed the identification of areas with incidence above the average for all analyzed pests, especially in the central area of the map. The process of evaluation described in this study can serve as an aid for risk analysis, promoting management benefits and reducing cost in the farms.