AI-guided optimization of nanocomposite-based phytoremediation for enhanced soil and crop health in contaminated land

Contamination of soil by multiple toxic metals requires remediation strategies that combine high removal efficiency, ecological compatibility, and rational materials optimization. Here, an explainable artificial intelligence (XAI)-assisted framework was developed to optimize multifunctional GO-biochar-PEI-metal oxide hybrid nanocomposites for coupled contaminant removal and phytoremediation. Random forest (RF) and artificial neural network (ANN) models were trained using independent physicochemical and operational descriptors to predict an Overall Performance Index integrating metal removal, plant biomass, and rhizosphere microbial response. RF outperformed ANN in the held-out test set ( \(\:{R}^{2}=0.9132\) vs. 0.5765) and five-fold cross-validation ( \(\:{R}^{2}=0.585\pm\:0.066\) vs. \(\:0.198\pm\:0.054\) ), demonstrating greater predictive robustness within the investigated design domain. Feature-attribution analysis identified nanocomposite dosage as the dominant design variable without assigning causal significance to model associations. Uncertainty propagation identified a model-predicted high-performance region near 350 mg kg − 1 . At this dosage, the GO-biochar-PEI-CuO hybrid achieved Pb(II), Cr(VI), and U(VI) removal efficiencies of 92.5%, 88.3%, and 85.7%, respectively, outperforming the corresponding TiO₂-containing hybrid. The enhanced performance is consistent with cooperative adsorption, coordination, porous retention, and photothermal-assisted interfacial processes, whereas possible redox transformations remain mechanistic hypotheses requiring direct speciation evidence. Given the simulation-assisted component of the dataset, broader generalizability requires independent experimental validation. Overall, this study establishes an interpretable AI-assisted materials-screening strategy for identifying robust operating windows and guiding the sustainable design of multifunctional hybrid nanocomposites for contaminated-soil remediation.