There is nothing wrong with using Perl for data analysis if you know what you are trying to do. There are a number of options for conducting statistical analyses from a classical POV. Bayesian methods are sadly lacking right now, but you can always call out to R for that.
If you just want to unleash algorithms on vast quantities of data (of unverifiable quality), Perl has some "machine learning" options, but they are limited, as are tutorials.
A better use would be to see how machine learning algorithms could improve Perl on systems that do not get a lot of testing. That is what
I am focusing on right now. I expect I'll need to write bindings to various C++ libraries, which is not the most appealing of options.