Novelty detection for time series data analysis in water distribution systems using support vector machines

摘要: The sampling frequency and quantity of time series data collected from water distribution systems has been increasing in recent years, giving rise to the potential for improving system knowledge if suitable automated techniques can be applied, particular, machine learning. Novelty (or anomaly) detection refers automatic identification novel or abnormal patterns embedded large amounts “normal” data. When dealing with (transformed into vectors), this means events amongst many normal points. support vector is a data-driven statistical technique that developed as tool classification regression. key features include robustness respect non-Gaussian errors outliers, selection decision boundary principled way, introduction nonlinearity feature space without explicitly requiring nonlinear algorithm by kernel functions. In research, regression used learning method anomaly flow pressure No use made past event histories through other information sources. methodology, whose derives training error function, applied case study.