<p>Massive Open Online Courses (MOOCs) are becoming more popular these days since they are affordable, flexible, and offer top-notch education to a large audience worldwide. However, due to a lack of timely support and direction from educators, students’ performance is not adequate. Distance also makes it challenging to evaluate MOOC students’ progress and offer prompt assistance. The MOOC performance prediction methods that are now available to us are unable to provide us with helpful prediction results or help us provide learners-focused intervention strategies. Moreover, their opinions were not finalized until the completion of the course, which caused a delay in the timely execution of corrective measures. Thus, to forecast students’ performance, this study suggests a day-wise multi-class model that makes use of the Echo State Network (ESN) and Deep Maxout Network. Within this structure, ESN extracted the students’ anatomical and behavioural characteristics and optimized the Deep Maxout Network utilized to forecast the students’ academic achievement based on these characteristics. Furthermore, an Auxiliary Classifier Generative Adversarial Network is employed to address the issue of imbalanced data. By using this paradigm, we may proactively detect pupils who are not performing well early in the course, which allows for rapid intervention and individualized support to improve overall student achievement. The Open University Learning Analytics Dataset (OULAD) and UCI student performance dataset will be used for the performance evaluation of the suggested method utilizing f1-score, accuracy, precision, and recall metrics. The results of the recommended methodology are contrasted with state-of-the-art techniques, indicating that the recommended framework is superior to existing methods.</p>