The current chapter examines the implementation of machine learning for water quality event detection. Two alternatives—supervised and unsupervised - are examined. It can be seen that the former performs better with the existence of historical classified true events, while the latter is preferable when history is not a factor. Water properties and their influence on methodology performance are also examined. The paper explains that TOC is preferred when false negative errors are rare or when experts' probability to make mistakes is low. The manuscript provides a numerical example that illustrates the two above-mentioned methodologies.
- False Alarm
- Receiver Operating Characteristic Curve
- United States Environmental Protection Agency
- Unsupervised Learning
- Water Utility
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.