My experience is that tests and risks are identified when new features are introduced or when defects are fixed. All this was done "manually" in all my works.
Your interesting idea is basically a type of CI server plug-in / component for focusing testing based on statistics of previous defects and analyzing the complexity of modified files.
You will obviously need a code / file map to check for cases, and if you had the opposite (failed test cases and files that were changed to fix), you would have some kind of automatic way to generate some information.
, "" "". .. "" , , - ... "" - , .