In many cases identifying exceptions or rare events can often lead to the discovery of unexpected knowledge. Outlier detection is used to identify these anomalous situations.
Such anomalies may be hard-to-find needles in a haystack, but may nonetheless represent high value when they are found (or costs if they are not found). Typical applications include fraud detection, identifying network intrusion, faults in a manufacturing processes, clinical trials, voting activities and criminal activities in E-commerce.
Applying machine learning to outlier detection problems brings new insight and better detection of outlier events. Machine learning can take into account many disparate sources of data and find correlations that are too obscure for human analysis to identify.
Take the example of credit card fraud: with machine learning online behavior (web site browsing history) of the purchaser becomes a part of the fraud detection algorithm – rather than simply considering the history of purchases made by the card holder. This involves analyzing huge amounts of data, but it also is a far more robust approach to E-commerce fraud detection.