Using software to automate the selection of relevant tests based on code changes. The goal is to reduce a large suite of automated tests to just the tests that are relevant to the change so that tests run quickly.
Many methods can be used to accomplish this. Presumably, Microsoft coined the term to describe automatic test selection in Visual Studio in 2011. Microsoft’s method is to analyze the code dependency for each test and use that to determine which tests to run. Paul Hammant has catalogued a number of different techniques for this on Martin Fowler’s website in his helpful article on The Rise of Test Impact Analysis.
Predictive Test Selection is one method of doing Test Impact Analysis with machine learning. Launchable’s adaptive subsets are an implementation of predictive test selection.