科学有效地预测水质对于水环境的可持续发展和人类健康具有重要意义，为此以固原市某黄河断面的水质监测数据为研究对象，提出了基于指标客观性的权重赋权(Criteria Importance Though Intercriteria Correlation，CRITIC)法和改进的秃鹰搜索(Improved Bald Eagle Search，IBES)算法优化双向长短时记忆网络(Bidirectional Long Short-Term Memory Network，BiLSTM)的组合水质等级预测模型。首先，采用CRITIC法确定各水质指标的权重，加权求和获得一项综合水质指标，从而提出一种改进的水质评价指标体系，以为BiLSTM提供更丰富、更可靠的水质特征信息。其次，在训练过程中引入Logistic映射和莱维飞行策略，并设计交叉共享及准反向搜索策略优化秃鹰搜索(Bald Eagle Search，BES)算法，以提升其种群多样性，增强寻优能力。最后，通过IBES算法迭代寻找BiLSTM的最佳学习率、隐藏层节点数以及正则化系数的超参数组合，进一步提高其预测水平。结果显示:与 IBES-BiLSTM、BES-BiLSTM、GA-BiLSTM、PSO-BiLSTM 和 BiLSTM 等模型相比，CRITIC-IBES-BiLSTM 模型进行水质等级预测的准确率、精准率、召回率及F1均最高，且具有更好的稳定性。
Research on water quality prediction using CRITIC and multi-strategy bald eagle optimization for BiLSTM
Accurate prediction of water quality levels enables us to detect abnormal trends promptly,allowing for targeted measures to protect water resources,preserve the ecological environment,and foster sustainable economic and social development.Hence,this study focuses on the water quality monitoring data from a specific section of the Yellow River in Guyuan City.It introduces a hybrid water quality level prediction model optimized by the Criteria Importance Though Intercriteria Correlation(CRITIC)method and an Improved Bald Eagle Search(IBES)algorithm in conjunction with a Bidirectional Long Short-Term Memory Network(BiLSTM).First,the CRITIC method was employed to analyze and compute the weights of the five indicators in the original water quality evaluation index system.Subsequently,the weighted sum method was utilized to derive a comprehensive water quality indicator,leading to the development of an enhanced water quality evaluation index system.This new system provides the BiLSTM model with more robust and comprehensive water quality characteristic information.Secondly,to improve the population diversity of the Bald Eagle Search(BES)algorithm and its capability to escape local optima and conduct global searches more effectively,a Logistic mapping strategy was incorporated into the population initialization process.Furthermore,the search space selection stage saw integration of the Levy flight strategy,and a cross-sharing strategy was introduced during the search space prey stage.Upon completion of the dive capture prey stage,a quasi reverse search strategy was devised.Finally,the IBES algorithm was employed to finely adjust the three key hyperparameters of the BiLSTM neural network,including the learning rate,number of hidden layer nodes,and regularization coefficient,thereby significantly enhancing the predictive ability of the BiLSTM model.Experimental results demonstrate that compared to IBES-BiLSTM,BES-BiLSTM,GA-BiLSTM,PSO-BiLSTM,and traditional BiLSTM models,the proposed model achieves superior performance with accuracy,precision,recall,and F1 values of 96.70％,94.48％,95.82％,and 95.04％,respectively,exhibiting strong stability.This outcome not only validates the effectiveness of the proposed model but also provides a robust foundation and technical support for further advancements in water quality grade prediction.
environmental engineeringwater quality predictionCriteria Importance Though Intercriteria Correlation(CRITIC)methodimproved bald eagle search algorithmBidirectional Long Short-Term Memory Network(BiLSTM)
万方数据
维普
