An interpretable machine learning model for predicting forest fire danger based on Bayesian optimization