Very important data mining (data analytical) pattern finding method. It is the basic method for intrusion / fraud detection and system health check. These all there areas needs to know about anomalies which are very different from other data points.

Very simple scholastic example – analysis of runtime of aggregations on database system:

# vector of runtimes in seconds
aggtime<-c(1.2, 2.1, 1.9, 2.3, 0.9, 2.8, 45) # basic statistics of vector > boxplot.stats(aggtime)
$stats
[1] 0.90 1.55 2.10 2.55 2.80

$n
[1] 7

$conf
[1] 1.502816 2.697184

$out
[1] 45

# shows that value 45 is the only outlier

We can draw a boxplot for these simple data:
anomaly01