Discovering Rules from a National Exam Repository: A Use Case for Data Analysis from Iranian Medical Schools Entry Exam.

Abstract

Many methods have been studied to analyze and interpret patterns and relationships that are embedded in the database to discover new knowledge in educational systems. Association rule mining is a type of data mining that identifies relationships among elements of the dataset. However, because these methods often generate various rules including non-significant ones, it is important to identify the most useful rules. Therefore, evaluating and ranking rules has become a topic of interest in the decision-making process in order to represent the level of usefulness of rules. We incorporated Apriori and Eclat algorithms on an educational dataset of a national medical exam in Iran. The aim of this study is to identify the usefulness of the extracted rules. This method can reliably discover new knowledge by interpreting the prioritized rules. The results show that those who have Scored in the highest category, i.e. [407,493], are accepted and who have scored in the lowest category, i.e. [150,236), are not accepted in the exam regardless of others features. Although, the rules that implication Accept=0 occurs, find out with high confidence, due to a large number of samples in this case. The ranking rules show this method is effective in the identification of insignificant rules that have no effect on decision making.