Detecting anomalies in external data
Anomaly detection is also considered to be an important application of unsupervised learning in general. Anomaly detection can be used in cases where you want to perform any kind of filtering of your existing data, called outlier detection, and also act as a real-time detector during the inference stage given new external data, known as novelty detection. Here are some examples of end user use cases of anomaly detection:
- Removing noise in your dataset that will be fed into a supervised feature learning process to enable more stable learning.
- Removing defective products in the production line. This can range from the manufacturing production of semiconductor wafers to egg production.
- Fraud prevention by detecting anomalous transactions.
- Scam detection through SMS, email, or direct messenger platforms.
Anomaly detection is a two-class or binary problem. This means that an alternative way people approach these example use...
