Multi-agent adaptive boosting on semi-supervised water supply clusters

The division of a water supply network (WSN) into isolated supply clusters aims at improving the management of the whole system. This paper deals with the application of spectral clustering to achieve this aim. A semi-supervised approach can take into account the graph structure of a network and incorporate the corresponding hydraulic constraints and the other available vector information from the WSN. Several of the disadvantages of these methodologies stem from the largeness of the most WSN and the associated computational complexity. To solve these problems, we propose subsampling graph data to run successive weak clusters and build a single robust cluster configuration. The resulting methodology has been tested in a real network and can be used to successfully partition large WSNs.