Water Quality Prediction Approach Based on t-SNE and SA-BiLSTM

Water quality prediction is a basic work in water resource management. High performance computing and communications (HPCC) provide conditions for accurate water quality prediction. Water quality data are time and index correlated. In the meantime, the water quality data at different time contribute differently to the prediction result. According to these characteristics of water quality data, a water quality prediction approach based on t-distributed stochastic neighbor embedding (t-SNE) and self attention- bidirectional long short term memory neural network (SA-BiLSTM) is proposed in this paper. First, according to the index correlation of water quality data, a water quality feature extraction algorithm based on tSNE is proposed to obtain the effective features. Second, according to the time correlation and time difference of water quality data, a water quality prediction model based on SABiLSTM is established to predict the water quality accurately. Finally, the proposed approach is tested in two actual water quality datasets: Victoria Bay and Tai Lake. Experimental results show that compared with the state-of-the-art approaches, the proposed approach can make full use of the characteristics of water quality data to obtain better prediction performance.