Machine learning prediction for academic misconduct prediction: an analysis of binary classification metrics

Abstract

Academic misconduct is unethical behavior in academic work. To sustain integrity culture and mitigating unethical conducts among higher education institutions community, the academic misconduct detection must be done at an earlier stage. Thus, this study attempted to provide a new empirical contribution with the analysis of binary classification performances metrics to describe the ability of machine learning in predicting academic misconduct. Four machine learning algorithms have been used namely generalized linear model (GLM), logistic regression (LR), decision tree (DT), and random forest (RF). Beside performances comparison, this paper presents the analysis of academic misconduct factors that were constructed based on demography and fraud triangle theory (FTT). The findings showed that all the four machine learning algorithms have obtained good ability in the prediction models with the accuracy at above 80% and below 20% of the classification errors. Rationalization from the FTT attributes has shown as the most important factor in GLM, LR, and DT. In RF, opportunity of FTT attributes have become the most important. Compared to FTT attributes, demography attributes were not providing much benefits to all the machine learning models but remain applicable at very low weight correlations.