Aquaculture Environment Prediction Based on Improved LSTM Deep Learning Model

Abstract

In aquaculture, there is always a potential risk of changing the water environment, hindering the growth of aquatic products, or even causing mass death, causing great damage to farmers. Therefore, it is vital to predict the quality of water resources early. A lot of methods have been introduced, including SVM, GM, RNN. These methods focus only on forecasting water quality in general, as well as fewer diversity of forecasting parameters, but do not focus on water characteristics in aquaculture. In this paper, we propose an aquaculture environment prediction based on an improved LSTM (long-short-term memory network) deep learning model. We conduct a characteristic analysis of the environmental parameters of lobster culture. Then use these features to improve the traditional LSTM model to improve the accuracy of the prediction model. The data used to train and test the proposed model are exploited from the actual set of environmental parameters measurement data for lobster farming of the environmental monitoring center in the Xuan Dai bay area, Phu Yen province, Vietnam. The prediction results of the improved LSTM model are compared with those of the RNN models. The results show that the improved LSTM model performs more accurate predictions of changes in aquatic environmental parameters than other compared solutions.