This notebook documents a series of experiments in anomaly detection on a time-series of measurements from a municipal water tank. It is inspired by the techniques described in Practical Machine Learning: A New Look at Anomaly Detection by Ted Dunning and Ellen Friedman, and realized in Python by Matthew Rahtz.
In Part One I define the problem and develop the anomaly detection technique. In Part Two I refine the anomaly detection technique by applying smoothing to the data. In Part Three I apply the technique to simulated anomalous data. In Part Four I develop a different model, based on simple histograms, that avoids some of the problems identified when testing the clustering model.