ANÁLISE DE PADRÕES PRODUTIVOS E CLIMÁTICOS DA SOJA EM GOIÁS POR MEIO DE TÉCNICAS DE CLUSTERIZAÇÃO

Abstract

Soybeans are consolidating their position as the main commodity in Goiás, however, regional analysis of the sector lacks tools that transcend state averages and reveal local patterns. This work applied machine learning techniques to identify productive and climatic clusters in the state, processing historical productivity and meteorological data using the Python language. The methodology included the pre-processing of seasonal data and the application of the K-Means clustering algorithm. The model, configured with five groups (k=5), was validated by metrics such as Silhouette and Caliński-Harabasz Index, showing consistent segmentation. The results highlighted a group with high technological performance, maintaining productivities above 3,470 kg/ha even under restrictive rainfall regimes, suggesting the intensive use of irrigation. The results indicate that the application of clustering algorithms is effective for extracting knowledge from complex databases, allowing automatic distinction between zones of natural suitability and zones of high technological management.