Título: A comparative analysis of machine learning techiniques to dry bean grain classification
Autores: Alison Dantas de Moura, Maria Beatriz Rodrigues Martins, Bruno Riccelli dos Santos Silva & Jose Wellington Franco da Silva
Resumo: This study presented an automated system for the identification of dry bean grains types, in order to replace manual classification with a faster AI-based method. To achieve this, machine learning (ML) algorithms were employed, including K-Nearest Neighbors (KNN), Random Forest (RF), Decision Tree (DT), Support Vector Machine (SVM), and XGBoost (XGB), along with evaluation metrics such as accuracy, precision, recall, and F1- score. The study utilized a dataset containing 13,611 samples, with sixteen (16) numerical attributes describing grain characteristics (e.g., area, perimeter, aspect ratio) and one (1) categorical attribute (Class). Grid Search(GS) was applied to optimize model hyperparameters, and the Wilcoxon statistical test was used to find the best approach. The results indicate that XGBoost was the most efficient model, demonstrating significant differences according to the Wilcoxon test. Regarding metric evaluation, SVM achieved the highest accuracy with a score of 0.8954. These findings reveal that bean classification automation can improve process standardization and reliability in agricultural production.
Palavras-chave: Agricultural informatics; Grain morphology; Predictive modeling.
Páginas: 8
Código DOI: 10.21528/CBIC2025-1191935
Artigo em PDF: CBIC_2025_paper1191935.pdf
Arquivo BibTeX:
CBIC_2025_1191935.bib