Beginning Anomaly Detection Using

type I 

error and type II error. These errors are used in hypothesis testing where you have 

a null hypothesis (which usually says that there is no relation between two observed 

phenomena), and an alternate hypothesis (which aims to disprove the null hypothesis, 

meaning there is a statistically significant relation between the two observations).

A 

type I error is when the null hypothesis turns out to be true, but you reject it 

anyways in favor of the alternate hypothesis. In other words, a false positive, since you 

reject what turns out to be true to accept something that is false. A 

type II error is when 

the null hypothesis is accepted to be true (meaning you don’t reject the null hypothesis), 

but it turns out the null hypothesis is false, and that the alternate hypothesis is true. This 

is a false negative, since you accept what is false, but reject what is true.

For the context of the following definitions, assume that the condition is what you’re 

trying to prove. It could be something as simple as “this is animal sick.” The condition 

of the animal is either sick or healthy, and you’re trying to predict if it is sick or healthy. 

Here are some definitions:

• 

True positive: When the condition is true, and the prediction is  

also true

• 

True negative: When the condition is false, and the prediction is  

also false

• 

False positive: When the condition is false, but the prediction is true

• 

False negative: When the condition is true, but the prediction is false

Putting them together, you can form what is called a