Learning Analytics and Academic Efficiency in STEM Algebra Courses: Predictive Indicators from ALEKS Adaptive Data

Abstract

Foundational mathematics courses represent a critical point for student persistence and success in STEM programs. However, debate remains regarding which usage indicators in digital learning environments most accurately explain academic performance. The present study examined the differential impact of time on platform and adaptive practice volume on performance in an Algebra I course among engineering students. A dataset of 567 students from the ALEKS adaptive learning system was analyzed using Pearson correlations, multiple linear regression, and binomial logistic regression. Results showed that the number of adaptive exercises completed was the strongest predictor of academic performance, explaining, together with time on platform, 50.2% of the variance in achievement. In contrast, accumulated time on platform demonstrated a substantially smaller effect. Furthermore, each additional exercise reduced the probability of low academic performance by approximately 5%. These findings empirically distinguish between passive exposure and active engagement, providing relevant evidence for instructional design, learning analytics, and retention strategies in engineering education. Adaptive practice emerges as a strategic component for strengthening performance in foundational STEM mathematics courses.