Beginning Anomaly Detection Using

 Anomaly  Detection

Anomaly detection is finding patterns that do not adhere to what is considered as 

normal or expected behavior. Businesses can lose millions of dollars due to abnormal 

events. Consumers can also lose millions of dollars. In fact, there are many situations 

every day where people’s lives are at risk and where their property is at risk. If your 

bank account gets cleaned out, that’s a problem. If your water line breaks, flooding 

your basement, that’s a problem. If all flights get delayed, that’s a problem. You might 

have been misdiagnosed or not diagnosed at all with a health issue, which is a very big 

problem that directly impacts your well-being.

Figure 

8-1

 is an example of an anomaly showing a rainbow-colored fish in the 

blueish fish family.

Figure 8-1.  An example of an anomaly

Chapter 8   praCtiCal Use Cases of anomaly DeteCtion

---

299

In business use cases, everything is centered around data, and anomaly detection is 

the identification of abnormal data points, events, or observations that raise suspicions 

due to the fact that they differ significantly from the data perceived as normal or typical. 

Many such anomalies can impact the business operations or bottom lines significantly, 

which is why anomaly detection is gaining a lot of traction in certain industries and 

many businesses are investing heavily in technologies that can help them identify 

abnormal behavior before it is too late. Such proactive anomaly detection is becoming 

more and more visible, and due to the new technologies developed as part of the AI 

revolution, this problem is also getting solved in ways never possible before.

Figure 

8-2

 is an example of the daily number of cars that cross the Golden Gate 

Bridge in San Francisco.

The kind of anomaly detection that can potentially help businesses depends very 

much on the kind of data collected as part of the business operations and the kind of 

techniques and algorithms used as part of the strategy to perform the anomaly detection.