Scholarlink

<p>Biochar has emerged as a promising adsorbent for removing antibiotics from contaminated water because of its high surface area, porosity, and diverse functional groups. However, optimizing biochar properties for effective adsorption remains challenging due to the complex interactions among its physical and chemical composition and the pyrolysis conditions during production. The current study conducted to address the knowledge gap by developing novel machine learning (ML) based framework to predict the adsorption efficiency of optimized biochar for commonly detected antibiotics which has not been explored previously through the integration ML model for the optimization of biochar characteristics. This study aimed to develop a machine learning (ML)-based framework to predict the adsorption efficacy of biochar for four commonly detected antibiotics: tetracycline, sulfamethoxazole, sulfadiazine, and ciprofloxacin. Data was collected through a comprehensive literature review spanning 2012 to 2025, using Google Scholar and Web of Science. Three ML models—support vector regression (SVR), random forest (RF), and neural network (NN) were applied to predict antibiotic adsorption based on biochar characteristics. Feature selection was refined through hierarchical clustering and model-based feature importance analysis. Model performance was evaluated using the coefficient of determination (R<sup>2</sup>) and root mean square error (RMSE). Contour plots were generated to assess interactions among key features, offering insights into the models’ predictions of antibiotic adsorption. The RF model demonstrated the best overall performance (R<sup>2</sup>: 0.93 for training, 0.45 for testing; RMSE: 204.48 and 136.23, respectively), while the SVR and NN models had lower test R<sup>2</sup> values of 0.39 and 0.15, respectively. Although the NN model had the lowest training RMSE (77.86), it exhibited the highest test RMSE (1138.33), indicating overfitting. The most influential features across models included surface area, elemental composition (C, H, O), and pyrolysis temperature. Interaction analysis revealed that biochar’s with carbon content below 70%, when applied at a concentration of 500&#xa0;mg L⁻<sup>1</sup> in wastewater, significantly enhanced antibiotic adsorption. This suggests that moderately carbonized biochar’s may possess more favorable surface chemistry or functional group availability for adsorption. This study concludes that ML, combined with engineered biochar, offers a data-driven strategy for predicting and optimizing antibiotic adsorptions supporting a cost-effective, scalable, and sustainable approach to wastewater treatment. Furthermore, integrating advanced ML models like RF can streamline biochar optimization by identifying key properties that maximize adsorption efficiency.</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

Biochar-Based Water Remediation: A Machine Learning Approach to Antibiotic Adsorption Prediction

  • 期刊论文

  • 发表日期：

    2025年10月4日

查看全文
  • Ghulam Mustafa Kubar,
  • Aftab Ali Kubar,
  • Kashif Ali Kubar,
  • Khurram Shahzad

摘要

Biochar has emerged as a promising adsorbent for removing antibiotics from contaminated water because of its high surface area, porosity, and diverse functional groups. However, optimizing biochar properties for effective adsorption remains challenging due to the complex interactions among its physical and chemical composition and the pyrolysis conditions during production. The current study conducted to address the knowledge gap by developing novel machine learning (ML) based framework to predict the adsorption efficiency of optimized biochar for commonly detected antibiotics which has not been explored previously through the integration ML model for the optimization of biochar characteristics. This study aimed to develop a machine learning (ML)-based framework to predict the adsorption efficacy of biochar for four commonly detected antibiotics: tetracycline, sulfamethoxazole, sulfadiazine, and ciprofloxacin. Data was collected through a comprehensive literature review spanning 2012 to 2025, using Google Scholar and Web of Science. Three ML models—support vector regression (SVR), random forest (RF), and neural network (NN) were applied to predict antibiotic adsorption based on biochar characteristics. Feature selection was refined through hierarchical clustering and model-based feature importance analysis. Model performance was evaluated using the coefficient of determination (R2) and root mean square error (RMSE). Contour plots were generated to assess interactions among key features, offering insights into the models’ predictions of antibiotic adsorption. The RF model demonstrated the best overall performance (R2: 0.93 for training, 0.45 for testing; RMSE: 204.48 and 136.23, respectively), while the SVR and NN models had lower test R2 values of 0.39 and 0.15, respectively. Although the NN model had the lowest training RMSE (77.86), it exhibited the highest test RMSE (1138.33), indicating overfitting. The most influential features across models included surface area, elemental composition (C, H, O), and pyrolysis temperature. Interaction analysis revealed that biochar’s with carbon content below 70%, when applied at a concentration of 500 mg L⁻1 in wastewater, significantly enhanced antibiotic adsorption. This suggests that moderately carbonized biochar’s may possess more favorable surface chemistry or functional group availability for adsorption. This study concludes that ML, combined with engineered biochar, offers a data-driven strategy for predicting and optimizing antibiotic adsorptions supporting a cost-effective, scalable, and sustainable approach to wastewater treatment. Furthermore, integrating advanced ML models like RF can streamline biochar optimization by identifying key properties that maximize adsorption efficiency.