$U^2RPSA$: Leveraging Unsupervised Learning for Water Main Decision Support
Track: Work in Progress
Categories: Water Systems, Fundamental Research on AI/ML methods for CI, Open Challenges on CI
Keywords: Water Pipe Breakage, Uncertainty, Bayesian Networks, Risk Prediction, Survival Analysis
TL;DR: Leveraging Unsupervised Learning for Water Main Decision Support
Abstract: The urban water supply system is crucial for city life, yet it remains vulnerable to a range of disruptions, particularly in densely populated areas. The motivation for this work is to design decision-support algorithms for early prediction of water main breaks and to prevent the potential damage to life and property. We present a comprehensive approach that uses statistical learning techniques and Bayesian networks involving three key steps: unsupervised learning of a Bayesian network structure to handle uncertainty in data and action outcomes, water main break risk prediction using machine learning, and survival analysis to estimate the probability of a pipe's longevity. Utilizing a publicly available dataset, we provide an initial evaluation of our approach showing that it outperforms a state-of-the-art model while providing a holistic understanding of pipe breakage dynamics for infrastructure maintenance.
Submission Number: 5
Loading