Depending upon the problem we are trying to solve and depending upon the use case there are multiple types of PD model. Since a PD model is a classification problem industry widely uses logistic regression in order to have high interpretability if there is any metric that shows shift. Otherwise for more advanced techniques like Light Gradient Boosting, XG boost , Random forest and decision tree. These are all models with high accuracy and less interpretability for which the regulators so not prefer them but these techniques improve model efficiency by greater extend than a logistic regression model.