Apriori-Based Association Rule Mining for Academic Performance Pattern Discovery and Decision Support in Universities
Higher education institutions generate large volumes of academic records every session, yet institutional analysis of this data is typically limited to course-level descriptive statistics that cannot reveal cross-course performance dependencies. This study addresses that gap by developing an Apriori-based association rule mining framework for discovering hidden academic performance patterns and result correlations in university student records, evaluated using support, confidence, and lift. Using 343,791 raw academic records obtained from a Nigerian public university, spanning 23 departments across 8 faculties, the data were cleaned to 331,842 valid records and transformed into 42,969 student-semester transactions, with each item representing a course-grade combination. The Apriori algorithm was applied with a minimum support of 0.5% and a minimum confidence of 60%, generating 289 frequent itemsets (221 one-itemsets and 68 two-itemsets) and five validated association rules. The discovered rules revealed strong course-to-course performance dependencies, including two high-lift failure associations (PHY107_F → PHY101_F, lift 60.41; CHM101_F → PHY101_F, lift 58.68) and three high-confidence associations among General Studies (GST) courses, the most reliable being GST1212_A → GST1213_A (confidence 68.49%). All five rules met the minimum support and confidence thresholds, and all had lift well above 1 (15.52 to 60.41). The results demonstrate that association rule mining can uncover actionable, non-obvious relationships in academic data that conventional descriptive statistics cannot reveal, and a prototype decision-support framework translating these findings into early alerts, curriculum-review recommendations, and policy briefs for academic advisers, curriculum planners, and university administrators is presented.
Download Paper