Water quality PH prediction study based on improved Ceemdan and RLNNA

In recent years, water quality Ph detection has gradually become a hot spot. However, the existing classical models in the field of water quality prediction, such as machine learning and deep learning, have insufficient feature extraction for water quality detection, which leads to poor prediction accuracy. Based on this, this paper proposes a BiLSTM model based on improved CEEMDAN and RLNNA reinforcement learning. After testing on the data from the National Statistical Yearbook, the results show that the model proposed in this paper has good results in the water quality Ph prediction study, and the goodness-of-fit R-square reaches 0.99893, which is extremely close to 1. In addition, compared with the classical algorithm, the model proposed in this paper improves the MSE, RMSE, MAE, MAPE and R-square.