An efficent water quality prediction and evaluation method based on the DBN-LSTM model

Abstract

Currently, water environmental issues have been severe. Water quality not only affects the ecological balance of nature but is also closely related to human lives. Therefore, an accurate water quality prediction and assessment can prevent water eutrophication, thus protecting the environment and improving people's quality of life. Previous studies have shown that the Long Short-Term Memory (LSTM) neural network has a good performance in time series prediction. In view of that, this paper proposes a neural network model that integrates the Deep Belief Network (DBN) and LSTM network for water quality prediction. The real data are used to predict water quality, and the DBN network is used to extract data features, which are then fed to the LSTM network. The proposed combined model is verified by experiments, and experimental results show that the proposed DBN-LSTM model can significantly improve prediction accuracy compared to the LSTM. The results indicate that the addition of the DBN network can improve the feature extraction and denoising performances on various datasets. The proposed water quality prediction model is compared with both single- and multiple-factor evaluation methods. The comparison results show that the DBN-LSTM model outperforms the LSTM model in terms of prediction accuracy, where it can be used in water quality monitoring management and risk assessment.