Scholarlink

<p>With the advent of the big data era, the field of machine learning is rapidly flourishing. Data modeling and prediction methods based on machine learning provide a new approach to various regression prediction problems. In view of the time series concerning the characteristics of water quality, this paper introduces the correlation of water quality features, so as to improve the accuracy and generalization ability of water quality prediction models. By employing the whale optimization algorithm (WOA) for optimization, a new water quality prediction model is established by combining the next-generation reservoir computing and WOA (WOA-NGRC). To explore the predictive performance of the WOA-NGRC model, BP neural network, LSTM, ESN, and NGRC models are also constructed for comparison. The predictive performance and stability of the model were evaluated. Water quality data from five major rivers in China, including the Hai River, Huai River, and Yellow River, were collected from the platform of the Ministry of Ecology and Environment of the People's Republic of China for the period from 2021 to 2024. Based on the correlations of features, water temperature, PH, and dissolved oxygen were selected as feature indicators for model construction. WOA was employed to adjust the model parameters, and the prediction results of the model were compared and analyzed. According to the prediction results, the overall predictive performance of the WOA-NGRC model was significantly higher than that of the other four models. Specifically, the <i>R</i><sup>2</sup> for water temperature could reach 0.986, which is a 29% increase compared to the LSTM model under the same sample conditions. For datasets with less pronounced seasonal cycles, such as PH and dissolved oxygen, the predictive performance of the WOA-NGRC model was slightly lower. However, these values are all above 0.8 and surpass those of the other four models. By extracting the correlations of features and leveraging the optimization and stable structure provided by WOA, the WOA-NGRC model achieves high precision and application value in predictive problems such as water quality forecast.</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

Next-generation reservoir computing water quality prediction model based on the whale optimization algorithm

  • 期刊论文

  • 发表日期：

    2025年3月27日

查看全文
  • Junyu Zhou,
  • Lijun Pei,
  • Zhiwei Zheng

摘要

With the advent of the big data era, the field of machine learning is rapidly flourishing. Data modeling and prediction methods based on machine learning provide a new approach to various regression prediction problems. In view of the time series concerning the characteristics of water quality, this paper introduces the correlation of water quality features, so as to improve the accuracy and generalization ability of water quality prediction models. By employing the whale optimization algorithm (WOA) for optimization, a new water quality prediction model is established by combining the next-generation reservoir computing and WOA (WOA-NGRC). To explore the predictive performance of the WOA-NGRC model, BP neural network, LSTM, ESN, and NGRC models are also constructed for comparison. The predictive performance and stability of the model were evaluated. Water quality data from five major rivers in China, including the Hai River, Huai River, and Yellow River, were collected from the platform of the Ministry of Ecology and Environment of the People's Republic of China for the period from 2021 to 2024. Based on the correlations of features, water temperature, PH, and dissolved oxygen were selected as feature indicators for model construction. WOA was employed to adjust the model parameters, and the prediction results of the model were compared and analyzed. According to the prediction results, the overall predictive performance of the WOA-NGRC model was significantly higher than that of the other four models. Specifically, the R2 for water temperature could reach 0.986, which is a 29% increase compared to the LSTM model under the same sample conditions. For datasets with less pronounced seasonal cycles, such as PH and dissolved oxygen, the predictive performance of the WOA-NGRC model was slightly lower. However, these values are all above 0.8 and surpass those of the other four models. By extracting the correlations of features and leveraging the optimization and stable structure provided by WOA, the WOA-NGRC model achieves high precision and application value in predictive problems such as water quality forecast.