Figure 3
From: pH trends and seasonal cycle in the coastal Balearic Sea reconstructed through machine learning
Bidirectional LSTM Neural Network model applied to fill the gaps in the pH\(_{T}\) time series in Cabrera: (a) training process monitoring loss for both training and validation sets, (b) Predicted pH\(_{T}\) values against their true values where the black line is the reference for perfect prediction, (c) Predicted pH\(_{T}\) time series in the training process (orange) and ground truth series (blue) and (d) Gaps in the pH\(_{T}\) time series filled with the trained model (red), while measured pH\(_{T}\) are shown in blue and predicted data in the training process shown in orange.
