Graph Convolutional Networks: Application to Database Completion of Wastewater Networks

Abstract

Wastewater networks are mandatory for urbanisation. Their management, including the prediction and planning of repairs and expansion operations, requires precise information on their underground components (manhole covers, equipment, nodes, and pipes). However, due to their years of service and to the increasing number of maintenance operations they may have undergone over time, the attributes and characteristics associated with the various objects constituting a network are not all available at a given time. This is partly because (i) the multiple actors that carry out repairs and extensions are not necessarily the operators who ensure the continuous functioning of the network, and (ii) the undertaken changes are not properly tracked and reported. Therefore, databases related to wastewater networks may suffer from missing data. To overcome this problem, we aim to exploit the structure of wastewater networks in the learning process of machine learning approaches, using topology and the relationship between components, to complete the missing values of pipes. Our results show that Graph Convolutional Network (GCN) models yield better results than classical methods and represent a useful tool for missing data completion.

Modeling land susceptibility to wind erosion hazards using LASSO regression and graph convolutional networks

“…The calculated AUC values for the training and validation datasets were found to be the same (97.2% and 97.25%, respectively), thus indicating an excellent prediction accuracy (Yesilnacar, 2005). The results show that GCNs significantly improve the performance of the predictions for land susceptibility to wind erosion and support the findings of previous studies, which suggested that GCNs are an efficient graph model for semi-supervised learning (Chen et al, 2018) and constitute a useful tool for completing missing data (Belghaddar et al, 2021). The DL hybrid models (e.g., convolutional neural network-gated recurrent unit (CNN-GRU) and dense layer deep learning-random forest (DLDL-RF)) were also recently shown to be efficient methods for classifying dust sources in the Middle East (Gholami and Mohammadifar, 2022).…”

Parametrization of a wastewater hydraulic model under incomplete data constraint

“…Using minimal and maximal value bounds chosen by the user, the algorithm allocates the pipe diameters according to Strahler's order [9], thus ensuring the general increase of the diameters from upstream to downstream. When a minimum of 20% of the attributes are available, one may use a semi-supervised learning method to predict the missing values as described in [10].…”

Machine Learning‐Based Surrogate Modeling for Urban Water Networks: Review and Future Research Directions

“…Therefore, these applications are not considered as surrogate models. In UDSs, Belghaddar et al (2021) applied GNNs to complete missing values in databases of wastewater networks.…”

“…In UDSs, Belghaddar et al. (2021) applied GNNs to complete missing values in databases of wastewater networks. Given its novelty and potential, further research on the GNN architecture is recommended to establish the benefits and limitations of this approach for surrogating UWN models, together with comparisons against already established MLSMs, for example, fully connected neural networks.…”