Predictive Classification of Water Consumption Time Series Using Non-homogeneous Markov Models

• DOI: 10.1109/dsaa.2017.32
• Publication Date: 2018-01-18T17:51:47Z
• Authors: Milad Leyli Abadi; Allou Same; Latifa Oukhellou; Nicolas Cheifetz; Pierre Mandel; Cedric Feliers; Olivier Chesneau
• Keywords: Categorical variable; Smart meter; Consumption; Water consumption; Predictive modelling

The analysis of time series data issued from smart meters has been studied relatively extensively in the electricity domain. Meanwhile, medium resolution water consumption collected via become possible recently, and research tried to develop statistical machine learning tools order respond different requirements domain, e.g., better understanding behaviors prediction consumption. In present paper, we propose a new predictive approach based on Non-homogeneous Markov Models learn dynamics behavior be able predict future with daily time-step relevant exogenous covariates. used for this purpose are categorical series, where each corresponds meter category specific behavior. experiments performed real set provided by utility France. Prediction results obtained proposed model compared those two models, namely, state independent homogeneous model. This classification can helpful utilities manage resources consumer requirements.