Abstract
Machine learning (ML) has employed various discretization methods topartition numerical attributes into intervals. However, an effectivediscretization technique remains elusive in many ML applications, such asassociation rule mining. Moreover, the existing discretization techniques donot reflect best the impact of the independent numerical factor on thedependent numerical target factor. This research aims to establish a benchmarkapproach for numerical attribute partitioning. We conduct an extensive analysisof human perceptions of partitioning a numerical attribute and compare theseperceptions with the results obtained from our two proposed measures. We alsoexamine the perceptions of experts in data science, statistics, and engineeringby employing numerical data visualization techniques. The analysis of collectedresponses reveals that $68.7\%$ of human responses approximately closely alignwith the values generated by our proposed measures. Based on these findings,our proposed measures may be used as one of the methods for discretizing thenumerical attributes.