Try Matlab http://www.mathworks.de/discovery/genetic-algorithm.html and its optimization toolbox if you are interested in genetic algorithms. It is better if you specify the type of problem rather than choosing a hammer without knowing the kind of nails...
consider WEKA, or RAPIDMINER then consider MATLAB or R. Nothing prevents you from using C or C++ if you are a good programmer and comfortable with these languages.
Im working with python to develop new methods and also multilabel learning problems due to weka havent support yet (Meka isnt good at all). There are basic libraries to work with (scipy, numpy) but there are complete frameworks for data mining and machine learning concretely (orange, mlpy, milk, pattern, scikit-learning and so on). For evolutionary computation i have used inspyred library. Python its free (GNU license) and has computational facilities to improve the performance over matlab and other script language with cython (C like implementation). [1,2]
If you are interested in multilabel learning im working on my own framework (mullpy) with automated capabilities for ensemble methods and graphic representations. Available soon.
[1] Prechelt, L. (2000). An empirical comparison of c, c++, java, perl, python, rexx, and tcl. Technical report, Fakultat for Informatik, Universitat Karlsruhe. Germany.
[2] Wilbers, I. M., Langtangen, H. P., and Odegard, A. Using cython to speed up numerical python programs. Technical report, Department of Informatics, University of Oslo.