<![CDATA[Academic stress classification is a significant challenge in education, as previous approaches often rely on opaque models or require large training datasets. This study develops a hybrid expert system for academic stress classification using forward chaining and Certainty Factor (CF) score fallback. The system was tested on 100 student cases with the following label distributions: Mild (48), Moderate (37), and High (13), classified independently by three experts. Label validity was tested using pairwise Cohen's kappa, yielding a mean value of 0.8280. The system achieved 100% accuracy, a 32% improvement over the classical forward chaining baseline (68%). Statistical evaluation using Wilson score intervals demonstrated high consistency across all key metrics (accuracy, precision, recall, F1-score) with a 95% CI of [96.4%, 100%]. The system is designed with an explicit and auditable rule structure, enabling deterministic classification based on symptoms. Although validation results are high, the unbalanced label distribution opens up the potential for spectrum bias. Going forward, the system is planned to be tested across institutions, assessed for integration with counseling services, and compared with other hybrid approaches. ]]>
Copyrights © 2025
