Scholarlink

<p>The present study aims to develop a machine learning-based framework for predicting and optimizing the photocatalytic degradation efficiency of methylene blue (MB) using TiO₂, Fe₂O₃, and Ti–Fe composite nanomaterials. A dataset comprising 706 experimental observations was compiled from literature sources, incorporating key reaction parameters such as catalyst type, catalyst loading, light intensity, solution pH, and reaction time. Various machine learning models, including linear, kernel-based, and ensemble methods, were evaluated for predictive performance. The results demonstrate that ensemble tree-based models significantly outperform conventional approaches in capturing the complex nonlinear relationships governing photocatalytic degradation. Among all models, Light Gradient Boosting Machine (LGBM) achieved the best predictive performance, which further improved after hyperparameter optimization, reaching a test R² of 0.8886 and RMSE of 9.71. HistGradientBoosting and CatBoost also showed competitive performance, while XGBoost demonstrated strong generalization capability with the highest cross-validation score. Feature importance analysis using Random Forest and SHAP revealed that reaction time is the most influential parameter, followed by light conditions and catalyst properties. The findings demonstrate that photocatalytic degradation is governed by complex nonlinear interactions among process variables, which ensemble models effectively capture. The observed trends are consistent with photocatalytic kinetics, in which prolonged irradiation enhances degradation efficiency until equilibrium is reached. Overall, this study highlights the effectiveness of machine learning in modeling and optimizing photocatalytic processes and provides a reliable data-driven framework for methylene blue degradation systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine Learning-based Prediction and Optimization of Photocatalytic Degradation of Methylene Blue Using TiO₂, Fe₂O₃, and Ti–Fe Composite Nanomaterials

  • 期刊论文

  • 发表日期：

    2026年5月14日

查看全文
  • Sweety Khut,
  • Aajinkya Singh,
  • Arpan Chatterjee,
  • Ankush B. Bindwal,
  • Pravin D. Patil,
  • Shamraja S. Nadar,
  • Manishkumar S. Tiwari

摘要

The present study aims to develop a machine learning-based framework for predicting and optimizing the photocatalytic degradation efficiency of methylene blue (MB) using TiO₂, Fe₂O₃, and Ti–Fe composite nanomaterials. A dataset comprising 706 experimental observations was compiled from literature sources, incorporating key reaction parameters such as catalyst type, catalyst loading, light intensity, solution pH, and reaction time. Various machine learning models, including linear, kernel-based, and ensemble methods, were evaluated for predictive performance. The results demonstrate that ensemble tree-based models significantly outperform conventional approaches in capturing the complex nonlinear relationships governing photocatalytic degradation. Among all models, Light Gradient Boosting Machine (LGBM) achieved the best predictive performance, which further improved after hyperparameter optimization, reaching a test R² of 0.8886 and RMSE of 9.71. HistGradientBoosting and CatBoost also showed competitive performance, while XGBoost demonstrated strong generalization capability with the highest cross-validation score. Feature importance analysis using Random Forest and SHAP revealed that reaction time is the most influential parameter, followed by light conditions and catalyst properties. The findings demonstrate that photocatalytic degradation is governed by complex nonlinear interactions among process variables, which ensemble models effectively capture. The observed trends are consistent with photocatalytic kinetics, in which prolonged irradiation enhances degradation efficiency until equilibrium is reached. Overall, this study highlights the effectiveness of machine learning in modeling and optimizing photocatalytic processes and provides a reliable data-driven framework for methylene blue degradation systems.