AI-Enhanced Learning Analytics and Student Performance Prediction at Higher Education Level
TL;DR
A theoretically informed AI-supported learning analytics framework in a pedagogically relevant way that goes beyond algorithmic optimization is developed and tested, and it shows good predictive accuracy, and that self-regulated learning behaviors proved to be important mediators.
Abstract
In the era of digital transformation, higher education institutions are producing enormous amounts of students' data, but many of them still do not fully utilize that data for educational insight to the benefit of student success. This gap is filled by this study developing and testing a theoretically informed AI-supported learning analytics framework in a pedagogically relevant way that goes beyond algorithmic optimization. The research design used was quantitative with a predictive correlational approach in which 1248 students from various institutes of higher education in Pakistan and abroad were involved. Data consists of LMS interaction logs, attendance records, assessment results, and self-regulation surveys that were validated. The proposed framework was tested using descriptive statistics, structural equation modeling and machine learning approaches for analyzing data. Results showed good predictive accuracy (82.4% for the identification of the at-risk students), and that self-regulated learning behaviors proved to be important mediators. The proportion of children with cycling, walking and active play increased significantly using AI-generated indicators, but demographic discrepancies remained. Practically, it recommends the universities how to implement in a responsible way that can support the value of teaching, student support, institutional decision-making, and ethical norms.