Trajectory Tracker: Crisis-Drift Detection System with AI-Powered Micro-Interventions to Support Student Retention

Student attrition in STEM programs represents a critical challenge, with dropout rates reaching 30-50% in undergraduate programs, translating to billions in lost tuition revenue. This paper presents a comprehensive Early Warning System(EWS) that leverages predictive learning analytics to identify at-risk students early in the academic semester. Our system introduces three key innovations: (1) Crisis vs. Drift detection that differentiates between sudden performance collapse and gradual disengagement, (2) AI-powered micro-intervention recommendations providing context-specific actionable guidance, and (3) transparent student-facing risk reports with clear explanations. Using a Random Forest classifier with calibrated probabilities, the system achieved perfect classification accuracy on training data. The interactive dashboard enables real-time risk assessment from week 4 onwards, empowering educators with timely, personalized intervention strategies. This student-centered approach addresses gaps in existing predictive learning analytics tools by emphasizing transparency, interpretability, and actionable outcomes for improved retention in higher education.