ExMed-BERT: A Transformer-Based Model Trained on Large Scale Claims Data for Prediction of Severe COVID Disease Progression

Code for ExMed-BERT training and fine-tuning.

Pre-trained model

Our pre-trained model is available at https://doi.org/10.5281/zenodo.7324178.

Relevant files

• scripts/calculate_iptw_scores.py: Script to calculate IPTWs
• scripts/pretrain-exmed-bert.py: Script to pre-train ExMed-BERT from scratch
• scripts/finetune-exmed-bert.py: Script to fine-tune ExMed-BERT for a classification task
• scripts/train-rf.py: Script to train an RF model for a classification task
• scripts/train-xgboost.py: Script to train an XGBoost model for a classification task

Contact

Please post a GitHub issue or write an e-mail to manuel.lentzen@scai.fraunhofer.de if you have any questions.