Prediction of attachment efficiency using machine learning on a comprehensive database and its validation