MACHINE LEARNING-BASED CLIMATE ANXIETY RISK PROFILING FOR UNIVERSITY STUDENTS: BASIS FOR CAMPUS ENVIRONMENTAL SUSTAINABILITY PLANNING

Abstract

Climate anxiety has emerged as an increasingly significant psychological concern among university students
due to the growing awareness of climate change and its associated impacts. However, conventional assessments
primarily rely on descriptive analyses and often lack predictive and intervention-oriented capabilities. This study
developed and evaluated an Intelligent Climate Anxiety Risk Profiling Framework for university students using
educational analytics, supervised machine learning, and explainable artificial intelligence. A quantitative
predictive analytics research design was employed using climate anxiety data collected from 315 undergraduate
students from Isabela State University–San Mariano Campus during the First Semester of Academic Year 2025–
2026. The Climate Change Anxiety Scale was utilized to measure four dimensions of climate anxiety:
Cognitive–Emotional Impairment, Functional Impairment, Personal Experience, and Behavioral Engagement.
A Climate Anxiety Index (CAI) was computed and transformed into low-, moderate-, and high-risk categories.
Three classification algorithms, namely Logistic Regression, Random Forest, and XGBoost, were developed
and compared using an 80:20 stratified train-test split and five-fold cross-validation. The results revealed that
the respondents exhibited a moderate overall level of climate anxiety (M = 3.254, SD = 0.685), with Behavioral
Engagement obtaining the highest mean score (M = 3.969). Most students were classified as moderate risk
(47.0%), followed by high risk (37.5%) and low risk (15.6%). Among the evaluated models, Logistic Regression
achieved the highest predictive performance, obtaining an accuracy of 88.9%, precision of 89.1%, recall of
89.3%, and F1-score of 89.2%. SHAP analysis identified Functional Impairment, Cognitive–Emotional
Impairment, and Personal Experience as the strongest predictors of climate anxiety risk. An operational example
further demonstrated how raw survey responses were transformed into interpretable risk profiles and evidence-
based intervention recommendations. The findings suggest that the proposed framework can serve as an
effective decision-support tool for the early identification of climate anxiety and the implementation of targeted
mental health interventions within higher education institutions.