Can Hamming loss be used as a fitness function in PSO for classification models?

Question

I am using PSO to optimize parameters for a classification model.

Since accuracy score and f1 score are not working well with the code, I am thinking of using hamming loss as the fitness function. Can this work?