Enhancing early detection of psychological disorders in children and adolescents using ML: a comprehensive mental health assessment

Enhancing early detection of psychological disorders in children and adolescents using ML: a comprehensive mental health assessment

Child and adolescent mental health present significant challenges with broad societal implications. However, early detection is often delayed, hindering timely interventions. This doctoral thesis aims to advance the field by integrating domain expertise with machine learning (ML) techniques to enhance predictive modelling and understanding of these conditions, enabling effective preventive strategies. Comprising five published papers and two submitted manuscripts, various facets of child and adolescent mental health are explored. Initial insights are derived from nationwide cross-sectional data provided by Young Minds Matter (YMM) study, focusing on children aged 4-17. Subsequent analysis using longitudinal data from Longitudinal Study of Australian Children (LSAC), which tracks 10,000 families, delves into early behavioural patterns. ML algorithms are applied to YMM dataset to identify optimal models for detecting mental health disorders such as depression, obsessive-compulsive disorder (OCD), separation anxiety disorder (SAD), and attention deficit hyperactivity disorder (ADHD). Significant features are identified using Boruta with Random Forest (RF). Strong performance in depression detection is demonstrated by RF, while Gaussian Naïve Bayes (GaussianNB) shows promise in OCD classification. RF proves effective in diagnosing ADHD and SAD. Beyond predictive modelling, association rule mining using the Apriori algorithm on YMM dataset highlights critical issues like school refusal and absenteeism across age groups, emphasizing the need for targeted interventions. Depression detection is further investigated by leveraging demographic data from YMM and longitudinal insights from LSAC. RF is employed for effective diagnosis of depression, complemented by insights from Logistic Regression (LR) into depression likelihood across age groups. A 17- year longitudinal study extends predictions of depression in children and adolescents using ML techniques, revealing patterns of occurrence, persistence, and progression across developmental stages. Enhanced Principal Component Analysis (EPCA) integrates domain knowledge, validated by methods like Partial Dependence Plots (PDP), improving model interpretability and predictive accuracy in child and adolescent mental health.

File reproduced in accordance with the copyright policy of the publisher/author/creator.