基于机器学习的碳基材料对水中四环素吸附预测研究

Prediction of Tetracyclines from Water by Carbon-based Material via Machine Learning

利用现有文献中的碳基材料对水中四环素的吸附量的数据,以机器学习为方法准确地预测了不同碳基材料在不同环境条件下对水中四环素的吸附量。其中梯度提升树(GBDT)对四环素的吸附量预测效果最好(R^( 2)>0.99)。比表面积和孔容积是决定碳基材料对四环素吸附量的最主要的特征。除pH与pH_( pzc)对吸附量的贡献为负外,其余变量对吸附量均为正贡献,即特征重要性越明显时,对吸附量的提升越有利。整体而言,四环素在碳基材料上的吸附是一个物理过程,受吸附剂的物理特性和环境因素影响较大,而受碳基材料的化学特性的影响较小。

In this study,a dataset was established with data collected from existing publications and employed to predict tetracyclines adsorption amount onto carbon-based material via machine learning.The analysis showed that GBDT algorithm obtained the highest score over others with R^( 2)>0.99.Specific surface area and pore volume are the determining factors to adsorption amount of tetracyclines.Solution pH and pH pzc negatively contributed to adsorption amount of tetracyclines,while other features positively contributed to adsorption amount of tetracyclines.The adsorption of tetracyclines onto carbon-based material is physisorption where physical properties of adsorbents and environmental pressures strongly affected adsorption,while chemical properties of adsorbents did not.