Testing in Data Science

Most Data Scientists tend not to be familar with a procedure known as Unit Testing within Software Engineering. Unit Testing really should be required when pushing new code to the repository. Unit Testing helps to see whether your algorithm or application is working as it should. Oftentimes, as applications become more and more complex as new versions are added one needs to ensure that the core services works as it should.

Why?

1. Check that your code does what you think it does?
2. Guard against show-stoppers when merging new code to the repo.

How?

1. Readable
2. Simple
3. Unique

Example

References:

Test-Driven Machine Learning by Justin Bozonier