Student CGPA Prediction and Academic Risk Analysis Using Machine Learning Techniques

Predicting one’s academic performance accurately helps educational institutes to take corrective steps before students’ performance falls irrecoverably. Existing systems used to predict the students’ performance mainly depend on either common behavioural factors that are difficult to capture reliably or on factors that are seldom related to academics. The suggested system considers only those academic factors that can be easily measured/collected in any educational institute namely hours of study, attendance percentage, number of backlogs, CGPA scores of midterm exams of all the subjects. A machine learning approach is used to develop the system which can predict students’ CGPA, program risk factor, pass/fail and also generate recommendation for improvement. The system employs a 2000-set synthetic dataset of academic records, created to ensure credible distributions of student performances within and across subjects. Dynamic contents of subjects have been implemented to enable predictions involving up to five subjects on different methods. Feature engineering techniques have been applied to obtain subject averages, indicators of cross-subject academic strength, and labels for pass/fail, dropout risks. Several machine learning models are used in the system to perform various classification or regression tasks. Random Forest Regressor is used for CGPA predictions, Logistic Regression for pass/fail classifications, and Random Forest Classifier for dropout risks. A web application based on Streamlit has been implemented to deliver an interactive platform providing an academic analysis dashboard with live prediction. The system is also integrated with visualization functions to compare midterm prediction performance visually. Tests showed the stable and expositive prediction behaviour of chosen academic parameters for institutional application. The proposed system integrates prediction, academic monitoring and advisory features in a single place.