Performance of Artificial Intelligence in Predicting Future Depression Levels.

Area of Science:

• Computer Science
• Artificial Intelligence
• Mental Health Technology

Background:

• Depression diagnosis is challenging with conventional methods.
• Wearable AI technology shows promise for identifying depression using motor activity data.
• Existing models often lack accessibility and impartiality.

Purpose of the Study:

• To evaluate simple linear and non-linear models for depression level prediction.
• To compare the performance of eight distinct machine learning models.
• To assess the utility of physiological features, motor activity, and MADRAS scores.

Main Methods:

• Utilized the Depresjon dataset containing motor activity data.
• Compared eight models: Ridge, ElasticNet, Lasso, Random Forest, Gradient Boosting, Decision Trees, Support Vector Machines, and Multilayer Perceptron.
• Employed physiological features, motor activity data, and MADRAS scores for prediction.

Main Results:

• Simple linear and non-linear models demonstrated effective depression score estimation.
• Complex models were not necessary for accurate prediction in this context.
• The findings support the feasibility of using accessible wearable technology.

Conclusions:

• Accessible wearable technology can be leveraged for effective depression detection.
• Development of impartial and efficient depression identification techniques is feasible.
• This approach offers potential for improved mental health monitoring and intervention.