In an era where data-driven decision-making has entered nearly every sector, education is no exception. Predicting student performance plays a vital role in educational settings, enabling educators to identify students at the risk of failure, personalize learning experiences and enhance overall academic outcomes. In the last five years, the use of deep learning techniques has gained significant traction in the area of student performance prediction, particularly for blended and online learning environments. As access to large-scale educational data continues to expand, recent advancements in deep learning have motivate the researchers to investigate hybrid deep learning methods that leverage the strengths of multiple techniques to enhance prediction accuracy. This review paper provides an in-depth survey of recent advancements in hybrid deep learning approaches for predicting student performance by synthesizing findings from research studies published over the past five years (2020–2024) using PRISMA guidelines. By examining the recent literature, this paper seeks to offer insights into the current hybrid predictive models and methods being applied for student performance prediction, the datasets and features used for modelling, the evaluation metrics to measure the performance of the models and the emerging trends and challenges in the area.