A robust hybrid data driven approach to model biochar yield in terms of biomass pyrolysis

Biochar yield prediction plays a critical role in optimizing pyrolysis processes and advancing sustainable biomass utilization. This study introduces a hybrid machine learning framework that integrates Decision Tree models with four optimization strategies including Tabu Search, Ant Colony Optimization, Evolutionary Strategies, and Batch Bayesian Optimization (DT-BBO). A curated dataset of 211 samples was preprocessed using Leverage outlier detection and normalization to ensure model robustness. Among all tested models, the DT-BBO approach achieved the highest accuracy, with an R² of 0.98, an MSE of 1.9, and an AARE% of 2.2%, outperforming the other optimization techniques. SHAP analysis revealed that pyrolysis temperature, residence time, and ash content were the most influential parameters governing biochar yield. Comparative benchmarking against previously published models confirmed the superior predictive capability and stability of the proposed framework. The results demonstrate that the DT-BBO model offers a scalable, interpretable, and high-performing solution for biochar yield prediction, contributing to methodological innovation and sustainable biomass valorization.