Multifactor prediction of sea water quality based on improved K-LSTM

Abstract

To solve the problem of multifactor sea water quality accurate prediction, a prediction model based on the integration of the improved k-means clustering algorithm and Long-short term memory (K-LSTM) was proposed. Firstly, the random forest method is used to fill the missing values for each factor in the dataset. Then, the improved k-means algorithm is used to denoise the water quality samples according to the multifactor time series of water quality. Finally, an LSTM neural network is used to establish the multifactor prediction model of water quality. In this study, the water quality data of five typical sampling points in Niuguwan Ecological breeding Bases from 2017 to 2021 were selected for training and testing. The RMSE of the model was reduced by 23%, and the MAE was reduced by 60%. The results showed that the water quality indexes of the five monitoring stations were improved, but the PH value showed a rising trend, the water quality was alkaline, and the total phosphorus content exceeded the standard. This model has completed the accurate prediction of sea water quality in Niuguwan ecological breeding bases, which provides a new idea for water environment protection.