Utilizing Decision Tree Analysis for Optimizing Test Item Selection in the Development of Mathematical Procedures in Measurements and Geometry

This study investigates the application of Decision Tree analysis to optimize test item selection in the development of mathematical procedures, particularly in the domains of Measurement and Geometry. Employing Machine Learning techniques, the research focuses on assessing the mathematical proficiency of seventh-grade students. A predictive model is constructed and assessed using a dataset comprising responses from 528 students, with analysis conducted using Decision Trees with a depth of three layers. Utilizing Python via Google Colaboratory, the study evaluates three test formats-essay, multiple-choice, and hybrid on Precision, Recall, F-measure, Accuracy, and Error metrics. Findings highlight the superior efficacy of the essay format, achieving an accuracy of 0.77 and an error rate of 0.56, meeting established criteria. Conversely, the multiple-choice format demonstrates comparatively lower effectiveness, with an accuracy of 0.62 and an error rate of 0.74. These results underscore the significance of employing Decision Tree Analysis for informed test item selection, particularly favoring subjective formats in optimizing mathematical proficiency assessments within measurements and geometry.