Open AccessRiver Dissolved Oxygen Prediction Based on Random Forest and LSTM
Author(s) -
Juan Huan,
Bo Chen,
Xian Xu,
Hui Li,
Ming Bao Li,
Hao Zhou
Publication year - 2021
Publication title -
applied engineering in agriculture
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.276
H-Index - 54
eISSN - 1943-7838
pISSN - 0883-8542
DOI - 10.13031/aea.14496
Subject(s) - random forest , mean squared error , mean absolute percentage error , mean squared prediction error , computer science , water quality , mean absolute error , artificial intelligence , predictive modelling , correlation coefficient , approximation error , feature (linguistics) , statistics , machine learning , mathematics , algorithm , ecology , biology , linguistics , philosophy
HighlightsRandom Forest (RF) and LSTM were developed for river DO prediction. PH is the most important feature affecting DO prediction. The model base on RF is better than the model not on RF, and the dimensionality of the input data is reduced by RF. RF-LSTM model is outperformed SVR, RF-SVR, BP, RF-BP, LSTM, RNN models in DO prediction.Abstract. In order to improve the prediction accuracy of dissolved oxygen in rivers, a dissolved oxygen prediction model based on Random Forest (RF) and Long Short Term Memory networks (LSTM) is proposed. First, the Random Forest performs feature selection, which reduces the input dimension of the data and eliminates the influence of irrelevant variables on the prediction of dissolved oxygen. Then build the LSTM river dissolved oxygen prediction model to fit the relationship between water quality data and dissolved oxygen, and finally use real water quality data in the river for verification. The experimental results show that the mean square error (MSE), absolute error (MAE), mean absolute percentage error (MAPE), root mean square error (RMSE), and coefficient of determination (R2) of the RF-LSTM model are 0.658, 0.528, 13.502, 0.811, 0.744, respectively, which are better than other models. The RF-LSTM model has good predictive performance and can provide a reference for river water quality management. Keywords: Dissolved oxygen prediction, LSTM, Random forest, Time series, Water quality management.