Wireless sensor networks play an essential role in modern agriculture, as they facilitate the monitoring of different variables that have an impact on crop yields. The successful operation of WSNs is highly dependent on their accurate deployment in the field, which requires proper modeling of radio wave propagation. In this study, we evaluate three path loss models obtained from machine learning: K-Nearest-Neighbors, Random Forest, and Decision Tree. The measurements were carried out on a cassava crop, one of Colombia's most important agricultural products. Compared to vegetation models, the use of ML allows for predictions with reduced error.