Prediction model of sparse autoencoder-based bidirectional LSTM for wastewater flow rate

A Predictive Adaptive Learning Method for Multivariable Time Series With Mooney Viscosity Prediction as an Application Case

“…BiLSTM (Bidirectional Long Short-Term Memory Network), an extension of LSTM with reverse computation, enhances correlation across temporal information. [14] Despite these advance-ments, RNN networks struggle to emphasize key time series information, thereby impacting prediction accuracy. [15] In contrast, transformer [16], diverging from the cyclic network mechanism of RNN, do not rely on prior state inputs.Instead, transformer analyzed complete time series inputs, leveraging the attention mechanism to comprehensively learn temporal relationships, rendering them adept at handling longterm dependencies.…”