/CompNet

Primary LanguagePython

CompNet

This is the source code of paper "Order-free Medicine Combination Prediction With Graph Convolutional Reinforcement Learning".

Overview

CompNet is an end-to-end model mainly based on graph convolutional networks (GCN) and reinforcement learning (RL). Patient information and drug-drug interactions knowledge are utilized to provide safe and personalized prediction for medication combination. CompNet is tested on real-world clinical dataset MIMIC-III.

Requirements

pytorch >= 0.4

python >= 3.5

Running the code

Data preprocessing

  1. download MIMIC data and put DIAGNOSES_ICD.csv, PRESCRIPTIONS.csv, PROCEDURES_ICD.csv in ./MIMIC-III/
  2. download DDI data and put it in ./MIMIC-III/
  3. run code ./process_MIMIC.py

CompNet

run main_CompNet.py

Cite

Please cite our paper if you use this code in your own work:

@inproceedings{wang2019CompNet,
  title="{Order-free Medicine Combination Prediction With Graph Convolutional Reinforcement Learning}",
  author={Shanshan Wang and Pengjie Ren and Zhumin Chen and Zhaochun Ren and Jun Ma and Maarten de Rijke},
  Booktitle={{CIKM} 2019},
  year={2019}
}