基于随机森林模型的造纸废水出水质量预测方法

Prediction Method for Effluent Quality of Papermaking Wastewater Based on Random Forest Model

为提升造纸企业废水出水质量预测的自动化水平,强化废水处理能力,研究采用增广矩阵对造纸废水原始数据实施升维处理,进而通过动态慢性特征分析方法确定最佳特征数量。在此基础上,通过随机森林模型对当前特征数量下的样本数据进行回归预测分析。为验证动态慢性特征-随机森林复合模型的有效性,统计分析了该模型的预测误差并与其他模型进行对比,发现动态慢性特征-随机森林复合模型预测结果的平均绝对百分比误差(MAPE)、均方根误差(RMSE)和决定系数R2分别为0.02、2.88、0.81,显著优于传统的慢性特征-支持向量回归复合模型,显示出了较为理想的造纸废水出水质量预测准确度水平,具有一定的应用价值。

In order to improve the automation level of predicting the effluent quality of papermaking wastewater and enhance the wastewater treatment capacity,this study adopts an augmented matrix to perform dimensionality enhancement on the raw data of papermaking wastewater,and then determines the optimal number of features through dynamic chronic feature analysis.On this basis,regression prediction analysis is performed on the sample data with the current number of features using a random forest model.To verify the effectiveness of the dynamic chronic feature random forest composite model,the prediction error of this model was statistically analyzed and compared with other models.It was found that the average absolute percentage error(MAPE),mean square root error(RMSE),and coefficient of determination R2 of the prediction results of the dynamic chronic feature random forest composite model were 0.02,2.88,and 0.81,respectively,which were significantly better than the traditional chronic feature support vector regression composite model.This showed an ideal level of accuracy in predicting the effluent quality of papermaking wastewater and had certain application value.