Extensive literature exists on the prediction of students’ performance in online learning. The studies construct features from online activities to be used as predictors for performance prediction. However, the studies focus more on behaviour compared to cognitive-based aspects. Students’ effort is a multidimensional construct that reflects both behavioural and cognitive engagement, encompassing student, course, and environmental factors that influence performance. Among its dimensions, intensity has shown strong associations with performance. Therefore, this study focuses on the intensity of students’ effort for performance prediction. Moreover, in predicting students’ performance, when dealing with data, the processes are prone to bias, including during constructing the features. This study proposed and explored generated feature as a new feature for effort intensity to mitigate the bias which utilized log file data and applies an average ratio with the maximum value method. This method is able to reveal students’ intensity of effort in learning and show a high AUC-ROC score in prediction starting from the middle learning phase, highlighting the feature’s potential for timely intervention. The proposed feature can improve the previous feature under certain settings, including a consistent gain in the final learning phase (AUC-ROC increasing from 0.91 to 0.93). Statistical Parity Difference (SPD) was preserved while the proposed method enhanced recognition of students who distribute effort across multiple activities rather than concentrating on a single activity. The results offer a promising avenue for advancing fairness or bias mitigation and enrich the existing students’ performance prediction studies.
Mohamad et al. (Sun,) studied this question.