This paper studies the application of Deep Neural Networks as classifiers for predicting student performance using data obtained from instructional management systems. Through its neural network architecture, this model captures complex relationships between student interactions and academic outcomes by classifying performance into High, Medium, and Low classes with high predictive accuracy. Validating the performance of the model is done through various evaluation metrics; there is accuracy, recall, and precision, showing an 85.42% test accuracy and a Mean Squared Error of 1.73; the best predictive reliability in terms of good discriminative power between the different categories of performance is further highlighted by the ROC curves. However, analysis of training and test losses shows overfitting in later epochs, pointing to the requirement of techniques like regularization for improving generalization. This research is illustrative of deep learning in educational data mining with actionable insights toward data-driven instruction. Findings from this research highlight the worth of neural networks in predictive analytics for education as a tool that can enable interventions targeted to the improvement of students’ performance.