STUDENT PERFORMANCE PREDICTOR &TRACKERUSINGMACHINELEARNING (ML)

Abstract

The rapid growth of digital education and virtual learning platforms has produced vast amounts of student-related data, including attendance patterns, assignment submissions, behavioral indicators, learning styles, assessment scores, and engagement metrics. Traditional academic evaluation systems are primarily reactive, focusing on outcomes rather than continuous performance improvement. As a result, students who begin to fall behind often remain unnoticed until the final stages of a course, reducing opportunities for timely intervention. This research proposes a Student Performance Predictor and Tracker based on machine learning algorithms designed to identify performance trends, predict academic outcomes, and provide personalized recommendations. The system leverages classification, regression, clustering, and deep learning models to analyze historical and real-time student data, helping educators improve academic planning and enabling learners to monitor their progress more effectively. The study incorporates supervised learning models such as Random Forest, Gradient Boosting, Logistic Regression, and Support Vector Machines to predict student outcomes such as pass/fail probability, expected grades, risk of underperformance, and learning progress trajectory. Unsupervised learning techniques such as K-Means clustering are used to group students into learning behavior categories (e.g., high performers, moderate learners, at-risk learners). Additionally, deep learning architectures like LSTM networks support performance forecasting based on temporal learning behavior patterns. The proposed system also features a continuous tracking module integrated with dashboards that visualize academic metrics, engagement levels, and personalized insights. To address dataset imbalance commonly present in academic settings, techniques such as SMOTE and Random Undersampling were applied to ensure reliable predictions for at-risk groups. Experimental evaluation conducted on publicly available educational datasets (including online learning platforms and institutional academic records) demonstrated that ML-based predictors achieved high accuracy, with Random Forest reaching above 92% classification performance. The LSTM model displayed strong forecasting capabilities, effectively predicting student improvement or decline trends over time. The findings confirm that machine learning can significantly strengthen student monitoring, allowing institutions to provide proactive academic support and improve learning outcomes. The proposed system can serve as a digital academic assistant capable of identifying weaknesses early, supporting personalized learning pathways, and enhancing both teaching and learning efficiency.