Título de la obra: Enfoques, Enfoques, Teorías y Perspectivas de la Ingeniería de Sistemas y sus Programas Académicos Editores/compiladores: Guillermo...
The research results presented in this document use machine learning algorithms, specifically unsupervised learning, to characterize the behavior of production, yield, market prices, and exports of cassava (Manihot esculenta) and yam (Dioscorea spp.) crops in Colombia between 2018 and 2023. To carry out this research, official data on the variables were used, which were obtained from official sources of information in the country. A dimensionality reduction process was performed using PCA (Principal Component Analysis), and subsequently, clustering algorithms (KMeans and DBSCAN) were used to investigate the regional heterogeneity of these crops. Internal validation metrics such as Silhouette Score and Davies-Bouldin Index were used to validate the results obtained, which supported the quality of the structure of the clusters obtained. The findings show varied geographical situations, with high-yield clusters linked to favorable agroecological conditions and areas focused on export, especially in the case of yams. KMeans facilitated the segmentation of large homogeneous areas, while DBSCAN identified small but important atypical areas. This methodology provided a valuable tool for guiding data-driven agricultural decision-making, technological interventions, and strategies for internationalizing the agricultural sector for these tubers in Colombia.