Enhancement of Knowledge Concept Maps Using Deductive Reasoning with Educational Data

Teacher-led lessons can successfully convey concepts depending on the teacher’s preparation. However, it is challenging to convey a specific concept in an online environment without a well-designed learning path to guide students. Learning paths allow students to backtrack the prerequisite content from a specific lesson in which they are weak or skip to related content in which they have a strong understanding, resulting in efficient learning. Knowledge maps, as the basis of personalized learning paths, can be generated from educational data by deriving prerequisite relationships between two knowledge concepts. We have aimed to enhance knowledge maps by adding the prerequisite relationships obtained by applying deductive reasoning to previous maps. Using test data from Company D, we first generated prerequisite relationships using the least absolute shrinkage and selection operator, random forest, and hidden Markov model for three datasets of the company. Next, we derived additional prerequisite relationships by applying deductive reasoning. The results showed that the knowledge maps of the three datasets had accuracies of 59%, 55%, and 84%, respectively, which were 3%, 10%, and 4% higher than those of the prior maps. As a result, “at risk” students can perform better using the enhanced knowledge maps by applying deductive reasoning.