Using Decision Tree Classification Algorithm to Predict Learner Typologies for Project-Based Learning

Abstract

As educational data in tertiary institutions are becoming huge, it is important to deploy Data Mining algorithms in discovering knowledge and improving academic quality. One fast course delivery approach or trend, constructivism in higher education is based on Learner prioritization in the learning process where a learner transforms information, constructs hypothesis and makes decisions using mental models. Similar learner groupings for project-based learning through hidden patterns extraction can aid Active Learning and Instructor Monitoring. In our previous paper, K-means clustering algorithm was used to group learners with similar scores in three assessments. In this paper, we built a classifier model using the J48 Decision Tree Algorithm for predicting learner groupings after getting class labels through the K-means clustering algorithm. This classifier will help in predicting future groupings of learners for the same course and attributes. The weka simulation for the classifier model gave a 99.9% ROC Area curve, which indicates a general performance of the model and a 96.19% of correctly classified instances. The Confusion Matrix has 80% of the members correctly classified. The classification model has an extremely low FP Rate of 2%, another indication of a high performance predictive classifier.