Machine learning in predicting whistle-blowing intention of academic dishonesty with theory of planned behaviour

Abstract

The COVID-19 pandemic and its aftermath have caused most higher educations to choose to implement remote learning as a new method of instruction and assessment. Nevertheless, remote learning has been criticized by having adverse impact on academic integrity. Whistle-blowing has been regarded as an effective mechanism in limiting such unethical behavior. Thus, the main objective of this study is to identify the influence attributes of whistle-blowing intention among university students. The effectiveness of the whistle-blowing attributes was observed in prediction models based on machine learning technique. This paper presents the fundamental knowledge on evaluations of tree-based machine learning algorithms namely decision tree, random forest, to be compared with logistics regression and gradient linear model. A rigorous evaluation reports are provided that includes the area under curve (AUC) as a supplementary metric to measure the model accuracy. Additionally, to provide a clearer insight on the whistle-blowing prediction models, the pattern of influences from the whistle-blowing attributes based on the adoption of theory of planned behavior (TPB) and demography are presented. The findings revealed that both TPB and demography attributes contain some degree of impressive knowledge for the machine learning to generate a good prediction result.