Utilising AI technique to identify depression risk among doctoral students

Area of Science:

• Educational Psychology
• Artificial Intelligence in Education
• Mental Health Research

Background:

• Doctoral students exhibit a higher prevalence of depression compared to other demographics.
• Research specifically targeting depression risk in this population remains limited and findings are not universally applicable.

Purpose of the Study:

• To investigate factors contributing to high depression risk among doctoral students.
• To develop an effective method for identifying doctoral students at risk of depression using AI.
• To propose evidence-based strategies for universities to prevent and intervene in doctoral student depression.

Main Methods:

• Utilized Random Forest algorithm to select 13 key features from 37 potential depression risk factors in doctoral students.
• Employed a Multilayer Perceptron (MLP) neural network for predictive modeling, achieving 89.09% accuracy on the test set.
• Analyzed data from the 2019 Nature Global Doctoral Student Survey.

Main Results:

• Identified overwork, poor work-life balance, and negative supervisor-student relationships as significant characteristics of doctoral students at risk of depression.
• Developed an accurate AI-driven model for predicting depression risk in this population.
• Created a detailed profile of doctoral students susceptible to depression.

Conclusions:

• AI methods can effectively identify doctoral students at risk of depression.
• Addressing factors like workload and supervisor relationships is crucial for improving doctoral student mental health.
• Universities can implement targeted strategies to support vulnerable doctoral students, enhancing overall well-being.