<![CDATA[Decision Support Systems (DSS) in smart farming require methodologies capable of representing the holistic complexity of agricultural ecosystems. This study proposes a DSS framework based on unsupervised machine learning, specifically K-Means clustering, to automatically segment land profiles using IoT sensor records. The dataset consists of 500 global sensor data points covering seven essential environmental variables: soil moisture, pH, temperature, rainfall, humidity, sunlight duration, and the NDVI index. Through Principal Component Analysis (PCA) for dimensionality reduction and Silhouette Score evaluation, the system successfully identified and mapped seven land profiles with distinct microclimatic characteristics. Cross-tabulation analysis further demonstrates the principal novelty of this DSS, namely its ability to classify land into "Safe Zones" (Clusters 0, 3, and 4), which are characterized by Mild disease status and are suitable for Soybean, Cotton, and Maize, as well as "High-Risk Zones" (Clusters 1, 2, 5, and 6), which consistently correspond to Severe disease status. These findings indicate that a DSS based on environmental clustering is substantially more effective for crop selection recommendations and disease prevention than conventional predictive approaches. Ultimately, this framework provides farmers with actionable insights to optimize productivity and minimize agricultural risk]]>
Copyrights © 2026
