This paper presents a literature review of modern methods for predicting academic performance and recommender systems in education. Key factors influencing student success, as well as data analysis approaches, including machine learning, ensemble methods, and neural networks, are examined. Special attention is given to challenges in processing educational data, such as data incompleteness, grading subjectivity, and the complexity of considering multiple factors. Various neural network architectures, their application in academic performance prediction, and the selection and adaptability of activation functions affecting prediction accuracy and model training are analyzed. Additionally, modern recommender systems utilizing hybrid filtering methods are explored. The paper concludes with a discussion of promising research directions aimed at improving model accuracy and adapting algorithms to heterogeneous educational data.