Predictive Modeling of Mood Using Functional Network Connectivity: Differential Impacts in Exercise + Smart and Control Groups

Accurate mood prediction is essential for understanding brain health dynamics and developing interventions, particularly among young adult populations such as undergraduate students. This study explores the use of static functional network connectivity (sFNC) measures from resting-state fMRI (rs-fMRI) data to predict a composite mood index reflecting anxiety, depression, and stress. An autoencoder was employed to capture nonlinear relationships among mood variables, followed by Gaussian process regression (GPR) to predict the composite index at baseline and post-intervention, as well as changes over time. The model demonstrated strong predictive performance, with <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R^{2}$</tex> values of 0.60 (baseline), 0.81 (post-intervention), and 0.64 (change). K-means clustering identified three baseline mood groups (low, medium, high) and two post-intervention groups, linking higher mood index values with increased stress, anxiety, and depression. Notably, the Exercise+Smart group showed greater baseline variability and improved mood post-intervention, shifting to a lower composite mood index cluster.