search

LuChang-CS / Chet

Codes for AAAI 2022 paper: Context-aware Health Event Prediction via Transition Functions on Dynamic Disease Graphs
https://ojs.aaai.org/index.php/AAAI/article/view/20380
19 stars 6 forks source link

Some problems in MIMIC-IV #5

Open lzh-nwpu opened 1 year ago

lzh-nwpu commented 1 year ago

Hello, I am very interested in your paper. When I used your source code to reproduce the experiment following the settings in the paper, the recall rate of the diagnostic prediction task on the MIMIC-IV dataset was very high, and the F1 evaluation index for the heart failure task was almost 0. But the same code works fine on the MIMIC-III dataset. May I ask how to adjust the code? In addition, with the hyperparameter settings you mentioned in the paper, the amount of parameters on the MIMIC-IV dataset is also somewhat different, May I ask the correct hyperparameter settings?

LuChang-CS commented 1 year ago

Hi lizehaodashuaibi,

Sorry I didn't get a chance to monitor my email in the last summer. And thank you for your interest. For the diagnosis prediction task, I just changed the task and hidden_size to task = 'm' and hidden_size = '350'. It works fine on my device. If you still have the same problem, it could be from the data.