Analysis and Prediction of Patient Mortality and Length of Stay

Usage

Data Download

Before extracting data, one need to request permission and download MIMIC III dataset from https://mimic.physionet.org/gettingstarted/access/ as csv.gz files

Load data

Before loading data, “LABEVENTS.csv.gz”, “CHARTEVENTS.csv.gz”, “OUTPUTEVENTS.csv.gz” should be saved in a folder named raw_data, and all loaded items would be saved in the raw_data folder.

python3 load_data.py 

Data cleaning

Before cleaning data, a folder named “remove_outlier” need to be created.

python3 clean_data.py 

To review Exploratory Data Analysis of Length of Stay and Mortality

Please refer to the following jupyter notebook files

• los.ipynb
• mortality.ipynb

Extract Baseline Feature Set and Customized Feature Set

Please follow and run the folloing jupyter notebook files in order

• process_spas.ipynb
• process_data.ipynb
• feature_selection.ipynb

Train MLNN for Short Stay/Long Stay/In-Hospital Mortality Prediction

Please follow the steps in the following jupyter notebook file

• NN.ipynb

Running GRU Model

The GRU model needs to take a parameter --target to specify predicting length of stay or mortality rate. It could be run like:

python3 interpolation_GRU.py --target 3 
python3 interpolation_GRU.py --target 7
python3 interpolation_GRU.py --target m