Detection of child depression using machine learning methods.

Area of Science:

• Child and Adolescent Psychiatry
• Machine Learning in Healthcare
• Mental Health Research

Background:

• Childhood depression has severe, long-term impacts on individuals, families, and society.
• Early detection and accurate diagnosis are crucial to prevent severe consequences.
• No prior research has utilized machine learning on a high-prediction dataset like Young Minds Matter for this purpose.

Purpose of the Study:

• Develop a machine learning model to predict depression in children and adolescents (ages 4-17).
• Evaluate and compare the performance of various machine learning algorithms.
• Identify family activities and socioeconomic factors contributing to depression.

Main Methods:

• Utilized the Young Minds Matter (YMM) dataset (Australian Child and Adolescent Survey of Mental Health and Wellbeing 2013-14).
• Employed Boruta algorithm with Random Forest (RF) for feature selection.
• Used Tree-based Pipeline Optimization Tool (TPOT) to select models, including RF, XGBoost, Decision Tree, and Gaussian Naive Bayes.

Main Results:

• Identified 11 key features for depression detection, including mood, interest, sleep, and concentration changes.
• The Random Forest model achieved 99% accuracy and 99% precision in predicting depression.
• The RF model demonstrated superior performance across all metrics and a rapid execution time of 315ms.

Conclusions:

• The developed Random Forest model offers a highly accurate and informative approach to predicting child and adolescent depression.
• This model outperforms other evaluated algorithms in predictive accuracy and efficiency.
• The findings highlight the potential of machine learning for early identification and intervention in pediatric mental health.