Introducing Dynamic Threshold Apriori-Based Approach for Question Prediction through Frequent Pattern Mining

Predicting frequent patterns in exam question papers presents a major challenge for students in the education domain. This study proposes a novel approach to exam question prediction based on frequent pattern mining, introducing a Dynamic Threshold Apriori (DTA) algorithm. Unlike the traditional Apriori method that employs a fixed minimum support threshold for all itemsets, DTA dynamically adjusts the support threshold based on the size of the itemsets, thereby improving the retention of meaningful larger combinations that might otherwise be pruned. This modification represents an algorithmic novelty aimed at enhancing pattern discovery efficiency. Additionally, while Apriori has been extensively applied in domains such as market basket analysis, our work pioneers its application in the educational context by adapting it for exam question prediction, thereby offering a new application domain novelty as well. Historical question datasets are processed to mine association rules, and the impact of varying support and confidence thresholds is analyzed. Comparative analysis demonstrates that DTA achieves higher prediction accuracy and better pattern retention than standard Apriori, offering valuable insights to students and educators, and establishing DTA as a significant enhancement for educational data mining tasks.