Adaptive and Predictive Learning Analytics for Early Intervention Using Skill-Will Profiling in Engineering Education

This study introduces a novel approach to adaptive and predictive learning analytics through the development of a skill-will profiling model aimed at early intervention in engineering education.This approach was tested using data obtained from 120 undergraduate students.Higher-order thinking skills were measured through multiple-choice assessments mapped to the revised Bloom's taxonomy levels-Apply, Analyze, and Evaluate-where Evaluate received a slightly higher weight to reflect its greater cognitive complexity.The skill score was computed as the weighted mean of these three levels, whereas the will score was derived from a weighted combination of attendance records and semester grade point averages (SGPA1 and SGPA2).Both the scores were normalized to a 1-10 scale for comparison.A rule-based classifier segmented learners into four quadrants-Q1 (high skill-high will), Q2 (low skill-high will), Q3 (low skill-low will), and Q4 (high skill-low will)-using threshold values based on the mean skill (5.30) and will (8.16)scores, enabling precise identification of diverse learner profiles.The quadrant analysis revealed that students in Q1 and Q2 demonstrated significantly higher academic success, whereas those in Q3 exhibited the greatest risk of failure.To enable predictive intervention, a logistic regression model was applied using skill and will scores as inputs to predict academic outcomes (Pass/Fail).The model achieved a robust performance with an accuracy of 87.5%, precision of 90%, recall of 94.7%, and F1 score of 92.3%.These findings highlighted the value of combining cognitive, behavioral, and academic features in the early identification of at-risk students.The proposed skill-will profiling model supports adaptive instruction and provides a scalable foundation for data-informed interventions in engineering education.