Optimizing Feature Selection for Binary Classification with Noisy Labels: A Genetic Algorithm Approach