Comprehensive Symptom Prediction in Acute Psychiatric Inpatients Using Wearable-Based Deep Learning Models: Development and Validation Study

Doi:10.2196/preprints.65994

Authors of source code : rirakang@gachon.ac.kr, gyom1204@gachon.ac.kr  keylee@gachon.ac.kr

Current version of the project: ver. 0.1

 

Prerequisites

PyTorch version 2.2

Python 3.7.16

 

How to use

We present the source codes as part of our research project to assist users with a basic understanding of computational analysis. For details on the context in which the code is used, please refer to the original article.

 

The repository contains the following key folders:

1. folder “Data_Processing”: Includes preprocessing scripts that handle sensor and nonsensor features, transforming them into a suitable input layer for the model.
2. folder “CNNGRU”: Contains the code for building the deep learning model used in our research.
3. folder “feature_importance”: Contains scripts to compute the permutation feature importance of the model, helping to identify which features most influence predictions.

 

Contact Information of the Original Article

• Minseok Hong. garam21th@gmail.com
• Yong Min Ahn. aym@snu.ac.kr

 

Contact Information of Code

• Ri-Ra Kang. rirakang@gachon.ac.kr
• Yong-gyom Kim. gyom1204@gachon.ac.kr
• KangYoon Lee. keylee@gachon.ac.kr