Toward an explainable public health intelligence to detect depression using mobile application usage
Abstract
The purpose of this research is to design a public mental health intelligence based on explainable machine learning. Historical data on user application usage behavior for four subsequent semesters from the Cybersecurity and Confidence in the Spanish Households National Survey were used, and an IT artifact was developed to detect depression. Historical use of mobile applications can partially predict user depression symptoms, and when we add sociodemographic data (gender, educational level, and age), our model performance reaches acceptable results. Finally, we implement post-hoc explainable algorithms at local and global levels, providing us with a detailed analysis of the variables that derive depression.
Recommended Citation
Sabzizadeh, Ehsan; Rello, Luz; and Urueña, Alberto, "Toward an explainable public health intelligence to detect depression using mobile application usage" (2024). MCIS 2024 Proceedings. 47.
https://aisel.aisnet.org/mcis2024/47