Mastering Laplace Smoothing in Naive Bayes: Avoiding Overfitting