预测造纸废水出水指标的随机森林建模方法

Random Forest Modeling for Prediction of Effluent Indices of Papermaking Wastewater

关键词:  废水处理过程  随机森林模型  出水指标  回归模型

Key Words:wastewater treatment processes  random forest model  effluent indicators  regression model

摘要:出水化学需氧量（COD）与出水固形物含量（SS）是评价造纸废水处理工艺好坏的重要指标。为了更好地对其进行预测，提出了一种基于随机森林（RF）模型的方法，并以R语言为工具进行回归预测。对比偏最小二乘（PLS）模型、支持向量回归（SVR）模型、人工神经网络（ANN）模型等常规预测模型，发现RF模型具有预测精度高，结果误差小，泛化能力好，调整参数少等优点。在对出水COD进行预测时，RF模型的相关系数r为0.7954，相比于PLS、SVR、ANN分别提高了8.88%、10.73%、14.68%。在对出水SS进行预测时，RF模型的相关系数r为0.8551，相比于PLS、SVR、ANN分别提高了15.43%、24.25%、30.79%。

Abstract:Effluent chemical oxygen demand (COD) and effluent suspended solid (SS) are important quality indicators of papermaking wastewater treatment process. To improve the prediction performance of these two indicators, a random forest (RF) model was proposed and implemented regressing forecasting using R. Compared with the conventional models including partial least squares (PLS), support vector regression (SVR), and artificial neural network (ANN), the RF model had the advantages of high prediction accuracy, small error, better generalization perforamce, and fewer parameters adjustment. In terms of the effluent COD prediction, the correlation coefficient r value of RF was 0.7954, which increased by 8.88%, 10.73%, and 14.68% compared with PLS, SVR, and ANN, respectively. In terms of the effluent SS prediction, the r value of RF was 0.8551, which increased by 15.43%, 24.25% and 30.79% compared with PLS, SVR, and ANN, respectively.