Response: 

In data mining, anomaly detection is referred to as the identification of items or events that do not conform to an expected pattern or other items present in the dataset. This is an uncommon behaviour or pattern in the data. 

Three types of anomalies can be categorized broadly. 

1. Point anomalies 
2. Contextual anomalies 
3. Collective anomalies 

A single instance of data is considered to be nomalous if it's too far off from the rest. One of the examples of a typical business use case is about detecting credit card fraud based on "amount spent." This is a point anomaly. 

When the ABNORMALITY is context-specific, then it is tagged as contextual anomaly. This type of anomaly is quite common in time-series forecasting RELATED datasets. One of the examples of a typical business use case is that SPENDING 100 USD on food every day during the holiday season is normal, however, it may be odd otherwise. Assume we have seen a spike in sales during Thanksgiving or Christmas vacation times, this may be genuine and expected. However, observing such a surge in a non-festive season could be anomalous. 

When a set of data instances collectively helps in detecting anomalies, then it is categorized under "collective anomaly". One of the examples of a typical business use case is that someone is trying to perform a financial transaction form a remote machine accessing a source or host unexpectedly where he/she does not have the authority to do so, an anomaly that would be flagged as a potential fraud attack.