Non-local Catboost-based Prediction for Postpartum Hemorrhage

The prediction of Postpartum hemorrhage (PPH) is one of the most important tasks for assisted parturients. However, the complex relationships among risk factors for PPH make it challenging to predict. Therefore, we propose a deep learning method, NonCat, which introduces a non-local attention mechanism, combined with CatBoost algorithm, to build predictive model for PPH severity. The experimental results showed that the NonCat model was superior to other classification models in predicting the severity of PPH, and its accuracy rate, recall rate, F1 score and accuracy were 96.20%, 96.54%, 96.20% and 95.88%, respectively.