An Unsupervised Cluster: Learning Water Customer Behavior Using Variation of Information on a Reconstructed Phase Space
Date of Award
Doctor of Philosophy (PhD)
Electrical and Computer Engineering
Povinelli, Richard J.
Brown, Ronald H.
Corliss, George F.
The unsupervised clustering algorithm described in this dissertation addresses the need to divide a population of water utility customers into groups based on their similarities and differences, using only the measured flow data collected by water meters. After clustering, the groups represent customers with similar consumption behavior patterns and provide insight into ‘normal’ and ‘unusual’ customer behavior patterns. This research focuses upon individually metered water utility customers and includes both residential and commercial customer accounts serviced by utilities within North America. The contributions of this dissertation not only represent a novel academic work, but also solve a practical problem for the utility industry. This dissertation introduces a method of agglomerative clustering using information theoretic distance measures on Gaussian mixture models within a reconstructed phase space. The clustering method accommodates a utility’s limited human, financial, computational, and environmental resources. The proposed weighted variation of information distance measure for comparing Gaussian mixture models places emphasis upon those behaviors whose statistical distributions are more compact over those behaviors with large variation and contributes a novel addition to existing comparison options.