Techniques such as machine learning for performance prediction in a group of students have become a vital aid in current education. The academic outcomes can be predicted using features like the student records, attendance, socioeconomic status of the student, behaviour, etc.; by using supervised learning algorithms like Logistic Regression, SVM, and Random Forests in this method, the educators can predict the above-specified features in an efficient manner. This approach is to apply the models for early detection of such students so that we can help them before it is too late or offer them a learning plan that will suit them. The goal of this review is to identify the existing approaches of supervised learning algorithms for predicting students’ performance: to outline their advantages and disadvantages and the results achieved. The review mentions that despite the potential of the current models like Random Forests, SVM, and Logistic Regression, there is still the question about model interpretation, model scaling, and making them ‘real-time’. The next research endeavours will be able to close these gaps to the incorporation of deep learning models, explaining AI, and multi-modal data acquisition. Avatar-based technologies could enhance the precision and flexibility of the prediction models helping educators make better decisions. However, data privacy and fairness issues will also need to be tackled in order to promote the right use of the technologies. In conclusion, there is apparently a lot of potential for machine learning in education that will lead to the best of learner outcomes.