search

PeterChe1990 / GRU-D

GRU-D, a GRU-based model with trainable decays for multivariate time series classification with missing values/irregular samplings
MIT License
118 stars 37 forks source link

Data set generation #1

Closed KarthikaKP closed 5 years ago

KarthikaKP commented 5 years ago

Hi, I am new in to machine learning.Can you please provide the code for generating the data required by GRU-D from PhysioNet Challenge 2012 dataset for predicting 4 tasks together.

Thanks in advance, karthika

PeterChe1990 commented 5 years ago

Hi Karthika,

We use only Training set A from PhysioNet 2012 Challenge Dataset.

For the labels of 4 tasks: In-hospital mortality: In-hospital_death == 1 in Outcomes-a.txt Length-of-stay less than 3 days: Length_of_stay < 3 in Outcomes-a.txt Whether the patient had a cardiac condition: ICUType <= 2 or not in $PID.txt Whether the patient was recovering from surgery: (ICUType mod 2) == 0 or not in $PID.txt

Please refer to the paper and the supplementary S3.2 for more details. Currently, we do not have cleaned iPython notebook scripts to share.

KarthikaKP commented 5 years ago

Thanks a lot for your reply.Will look in to it and mail you if I have any doubt.

Thanks, Karthika