This study compared the predictive ability of gold-standard logistic regression (LR), decision tree (XGBoost: XGB) and deep learning (artificial neural network: ANN) algorithms to predict individual cases of anxiety, attention deficit, depression and disruptive behaviors in a large (n = 1120) sample of youth aged 5-21 y and their parents from the Healthy Brain Network (HBN) cohort. The study used an innovative AI meta-learning method to optimize algorithmic performance by jointly learning hyperparameters and performing automated, principled feature selection while also rendering deep learning interpretable for translational applications. Results showed that deep learning optimized with AI outperformed decision-tree and logistic regression in constructing individual-level psychiatric cases of major mental illnesses during the peri-adolescent developmental life stage.
Related Posts
