Student Performance Analysis to Identify the Students at Risk of Failure

One of the most important qualities of a developed civilization is its educational structure. Higher education institutions often struggle to determine why their pupils failed or what factors contribute to such outcomes. Predicting student performance remains one of the most significant subjects for learning environments like universities, as it aids in the development of effective strategies that enhance academic outcomes and prevent dropouts. These are aided by the automation of many procedures involved in typical student activities, which manage huge amounts of information collected through software applications for technology-oriented learning. Predicting student performance remains one of the most significant subjects for learning environments like universities, as it aids in the development of effective strategies that enhance academic outcomes and prevent dropouts, among other things. Thus, properly evaluating and processing these data might provide us with significant details about the students' knowledge as well as the link among them and academic performance. This data feeds potential algorithms and methodologies that can forecast students' success. In this paper, a smart framework is developed to forecast student failure at an academic level using online University learning data. The prediction framework is driven by machine learning algorithms, which use feature engineering, feature analysis, and resampling techniques to create a framework for identifying students at risk of failing a course. We further validate our framework through the K-Fold cross-validation technique. We tackle this challenge through two perspectives: a triple-class problem and a binary-class problem. Experimentation shows that our framework achieves excellent results with respect to student performance prediction with an accuracy of 90% and an F1 score of 0.90 for triple-class classification. In comparison, it achieves an accuracy of 94.8% and an F1 score of 0.95 for binary class classification.