The growing complexity of higher education demands proactive strategies to enhance student retention and academic performance. This paper presents an innovative approach to identifying students at risk of academic failure by leveraging a deep learning predictive model. The proposed model, based on a Multi-Layer Perceptron (MLP) neural network, analyzes historical student data, including socioeconomic and academic performance indicators. The primary objective is to predict the need for academic tutoring, enabling timely and targeted interventions. The methodology involves a meticulous data preprocessing pipeline, including data merging, cleaning, and normalization, followed by the training of the deep learning model with 100 epochs. The model’s performance is evaluated using metrics such as accuracy and loss, demonstrating its efficacy in this predictive task. The results of the experiment, which included predictions for 106 subjects, show that the model achieved an average accuracy of 0.7667. While this average performance reflects current processing capabilities and the diversity of the subjects, it is noteworthy that in several individual cases, the model achieved much higher accuracies, with the minimum loss recorded at 0.0528, indicating an exceptionally precise prediction of tutoring needs for specific subjects. Our findings indicate that deep learning models can serve as a powerful tool to complement human academic support systems, optimizing resource allocation and improving overall student success. This research contributes to the growing body of literature on the application of artificial intelligence in education, specifically in the context of academic support and student retention.