Carmen

AAAI-23 paper: Context-aware Safe Medication Recommendations with Molecular Graph and DDI Graph Embedding This is an implementation of our model Carmen and the baselines in the paper.

Requirements

torch == 1.8.0+cu111
torch-geometric == 1.0.3
torch-scatter == 2.0.9
torch-sparse == 0.6.12

Process Data

The processed data is in the path

\Carmen\data

You can also process data with

MIMIC-III

python processing.py

MIMIC-IV

python processing_4.py

Train Model

\Carmen\src

python main_train.py --cuda 2 --datadir ../data/ --MIMIC 3 --model_name Model_mimic3_DDIenc --encoder main --seed 312 --ddi_encoding --gnn_type gat --num_layer 10 --p_or_m minus

Test Model

python main_train.py --cuda 2 --datadir ../data/ --MIMIC 3 --model_name Model_mimic3_DDIenc --encoder main --seed 312 --ddi_encoding --gnn_type gat --num_layer 10 --p_or_m minus --Test --resume_path saved/Model_mimic3_DDIenc/best.model