Vegetation is needed to improve soil slope stability. The roots of different species stabilize the ground by their tensile strength. However, how the tensile strength is governed by different root and shoot characteristics is less known. In this study, root tensile strength was investigated, and root and shoot characteristics were simultaneously measured. An experimental relationship between the root tensile strength and root diameter was developed. First, feature selection methods were applied to identify the critical characteristics affecting root tensile strength. Furthermore, the machine learning (ML) models were developed using individual methods (Sequential Minimal Optimization Regression (SMOreg), Instance-based Learning (IBk), Random Forest (RF), Linear Regression (LR), Multi-layer Perceptron (MLP)) and ensemble methods (ensemble via MLP and LR models). Results showed that the ensemble via the MLP outperformed all other individual models as well as the ensemble via the LR. The root mean square error of the ensemble via MLP was 3 times better compared to the experimental model based upon a power relationship between root tensile strength and the root diameter. The need for studying the complexity in relationships of various attributes of vegetation using ML models is discussed.