APPLICATION OF ITEM RESPONSE THEORY IN ASSESSING THE RELEVANCE OF ELEMENTS IN CLUSTERS

Abstract

The objective of this study is to apply Item Response Theory (IRT) to determine the difficulties in assigning elements to their respective groups of 38 clustering algorithms in 11 datasets. Using the cognitive scale based on the Difficulty b-Parameter of the IRT Logistic Parameter Models (LP), results were obtained indicating significant changes in the relevance of elements in balanced and unbalanced groups, where the level of accuracy in assigning elements in the difficult range was predominant for the clustering algorithms. The use of the psychometric scale during grouping assignments in adverse situations increases the reliability of decision support systems, with potential use by professionals in critical areas.