- Problem common sighted in unsupervised learning as well as supervised learning
- Suppose we have set of features that each of them is going to test the aircraft engine.
- In the example above we just have two set of features that test the engine, so far engines worked well based on the given red mark
- Then we add new engine and make two feature to that engine. It turns out that the test reveal the engine being far away from the rest of the engine.
- This is what known in Machine Learning as Anomaly Detection.
- p(x) is the probability of features
- sigma in here denotes some threshold value
- Here we know checking how engine that we tested is less or more than particular threshold value(discussed later, computed by the Gaussian Distribution)
- fraud detection user example
- x's = denotes how much user login, typing speed, activity, dll
- technique used by web to analyzed human behaviour
- Manufacturing like jet plane example earlier
- AD on monitoring computers. if one of features is behaving strangely, then there's must be something wrong for the computer
- Next Gaussian distribution and gaussian algorithm for the tools that used for checking anomaly detection