Prediction of Online Academic dishonesty Using the Voting Ensemble Machine Learning Method

Due to pandemics, remote tests have grown in popularity, and with the benefit of remote working circumstances, the majority of enterprises and institutions employ online platforms for assessment. Despite the high degree of test monitoring, it is extremely simple for a student to acquire third-party assistance during the exam. To defend these sorts of operations, there are relatively few models that can anticipate the zone where the likelihood of fraud is greatest. In this research, we describe an academic dishonesty detection method for online tests that is based on machine learning. This algorithm predicts if a student is cheating or acting ethically based on a study of exam data performed after the exam. In this study, we analyze a large variety of variables for predicting unethical behavior during tests, including exam mode, number of questions completed within ideal time, observed time, accurate questions within ideal time, pass criteria, likelihood of question recurrence, and exam traversal. However, for initial modelling, only eleven elements are used. To increase the accuracy, we include all remaining components. The model predicts suspicious behavior when the number of right tries within the allotted time exceeds the threshold established based on the total number of exam questions. The link between time-based parameters, accuracy, and pass criteria gov-erns the machine learning model used to forecast fraud and real situations. Combining the findings of Extreme Random Forest, Logistic Regression, Random Forest, and XGBoost Classifier models, the suggested solution employs the “Voting Ensemble” approach of machine learning to increase the accuracy of prediction.