NoakLiu/CRRT-Medical-Treatment-Prediction-by-Applying-Reinforcement-Learning-Methods

This is a research project conducted by researchers from UW-Madison, Tsinghua University, Peking University People's Hospital

CRRT-Medical-Treatment-Prediction-by-Applying-Reinforcement-Learning-Methods

This is a research project when I worked as an internship in by Tsinghua University, cooperated withPeking University People's Hospital.

The contribution can be summarized as follows:

1. We extracted data by using PostgreSQL from MIMIC IV.

2. We realized the data interpolation by using the SAITS model.

3. We build two prediction model for CRRT treatment:

   3.1 LSTM+Conv model

   3.2 ”CTRL“ Self Defined Model

   ​ 3.2.1 UCB

   ​ 3.2.2 Regression Model

   ​ 3.2.2 Delayed Reward

   ​ 3.2.3 Bad Action Reward Decreasing

   ​ 3.2.4 Vital Sign Group Reward