Investigating and machine learning predicting the Impact of Nandina domestica Roots on the Unconfined Compressive Strength of unsaturated silty clay

Bio - stabilization of soil using natural vegetation offers a sustainable and cost-effective approach to slope management and erosion control. Furthermore, soft computing technologies are necessary since experimental research is often challenging, and time consuming. This article discusses the modelling and the performance of nonlinear and Random Forest (RF) model for predict UCS ultimate stress and trajectory of root-soil stress-strain. Nandina domestica is particularly promising due to its hardiness and low maintenance requirements, making it an economical and sustainable option for soil bio-stabilization. Using compressive strength, the effects of moisture content, root weight density (RWD), and root diameter (RD) on soil reinforcement were evaluated. 144 silty clay soil samples reinforced with four RWD (1, 2, 3 and 4 gr/cm ³ ) and three RD (1, 2 and 3 mm) in four moisture conditions (20, 25, 30, and 35%) was considered to determine the UCS. The proposed nonlinear model demonstrated acceptable predictive accuracy (R² = 0.84) with RMSE=8.7 kPa and MAE=7.13 kPa values predicted the contribution of Nandina domestica roots to soil ultimate UCS reinforcement. The data was split into three sections for the development of the RF model: the training data set (70%), the testing data set (15%) and the testing unseen data set (15%). The proposed model accurately predicted the trajectory of unknown data root-soil stress-strain and contribution of Nandina domestica roots to soil reinforcement, exhibiting strong performance with R² = 0.97, RMSE = 3 kPa, and MAE = 1.8 kPa.