Madbhavi, RahulRahulMadbhaviJoshi, AmitAmitJoshiMunikoti, SaiSaiMunikotiDas, LayaLayaDasMohapatra, Pranab KumarPranab KumarMohapatraSrinivasan, BabjiBabjiSrinivasan2025-08-312025-08-312020-10-02[9781728150703]10.1109/GUCON48875.2020.92311482-s2.0-85096596663https://d8.irins.org/handle/IITG2025/23977Leaks in water distribution networks (WDNs) contribute significantly to losses incurred by water utilities. Detecting and locating leaks is therefore essential to reduce these losses. This article proposes a framework and two feature selection methods for leak localization. These methods utilise the underlying correlation structure of the data to determine the set of least correlated features. The second method assigns weights to features by assigning more importance to features having more variability. The methods are evaluated by applying them on data generated by hydraulic simulations of three WDNs. Classification accuracies higher than 99.5% was obtained for the Hanoi and Fossolo networks with as few as four sensors and localization error was reduced by approximately 58% as compared to the best-case error reported in the literature.falseLeak localization | Machine learning | Water distribution networksConference Paper95-1002 October 202029231148cpConference Proceeding1