I just had an idea that's going to be simple to implement.

Step 1 - formats the perl code with Perl::Tidy and write to a temp file

Step 2 - compute the difference between the original code and the generated code, and generate some sort of index similar to the Levenshtein distance (Text::Levenshtein)

Step 3 - the idea is that the smaller the distance/difference to the style, the closer the code is to a particular style.

Otherwise good luck with your quest.