Norm-Referenced Achievement Grading of Normal, Skewed, and Imperfectly Normal Distributions Based on Machine Learning versus Statistical Techniques

Grading is the process of interpreting learning performance to inform learners and instructors of current learning ability levels and necessary improvement. In norm referenced grading, the instructors conventionally use heuristics, statistic methods, or the mix of both. In the rise of artificial intelligence, we investigate if a machine learning technique is reasonably helpful for norm referenced grading in comparison to a statistical method and a heuristic method. We conduct all of the grading methods based on a normal-distribution data set, skewed-distribution data sets, and imperfectly-normal-distribution data sets. The results reveal that heuristic method and machine learning method yield similar grading results in several cases. Based on the real data sets, heuristic, machine learning, and statistical methods produce the best, moderate, and lowest grading qualities, respectively, in terms of Davies Boulden index.