Predicting Textbook Media Selection Using Decision Tree Algorithms

Abstract

In this paper, we will investigate machine learning algorithms in predicting the medium of reading whether it be a digital or paper based. Synthetic Minority Oversampling Technique (SMOTE) filter is applied to oversample the dataset without affecting the original data and to find the optimal accuracy of algorithms. Four decision tree algorithms are examined (J48, Random Forest, Random Tree, and Rep Tree) over the data before and after applying SMOTE filter based on the performance criteria (TP rate, FP rate, Precision and Recall). Random Forest proved its accuracy with (0.743) for precision and (0.733) for recall over the other three algorithms.