Authors: Venugopal S, Dr. P. Jeyanthi
Abstract: Predicting student performance has grown in importance as a field of study in education analytics and machine learning. Large volumes of student data, including attendance, grades, assignments, and behavioral records, are produced by educational institutions. Conventional assessment techniques frequently fail to spot weak students early on. A machine learning-based method for forecasting students' academic success is presented in this research. In order to handle missing values and normalize attributes, the system gathers and preprocesses student data. The most pertinent elements influencing student performance are found using feature selection strategies. Prediction is done using machine learning algorithms like Decision Tree, Random Forest, SVM, and ANN. Students are divided into three performance categories by the trained models: High, Medium, and Low. Additionally, the suggested approach produces likelihood scores for predicting academic performance. According to experimental findings, ANN and Random Forest outperform conventional techniques in terms of prediction accuracy. The approach assists teachers in identifying kids who are at danger and in promptly offering academic support. Additionally, it helps schools raise student success rates and overall educational quality.
DOI: https://doi.org/10.5281/zenodo.20416403
