Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{179603,
        author = {Harish Kumar and Amit Singh Thakur and Ashok Kumar Behera and Chaitali Choudhary},
        title = {Predicting Student Dropout Rate from Institutions using MLP-BranchNet},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {11},
        number = {12},
        pages = {8337-8344},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=179603},
        abstract = {Student dropout from higher education institutions is a persistent challenge that affects academic performance, institutional reputation, and overall student success. Early identification of students at risk of dropping out can empower educational institutions to provide timely interventions. In this study, we propose a novel deep learning architecture named MLP- BranchNet, which leverages a hybrid branching structure based on Multi-Layer Perceptrons (MLPs) to improve the accuracy and explainability of student dropout prediction. The proposed model is trained on an institutional dataset after undergoing comprehensive steps including data cleaning, exploratory data analysis (EDA), preprocessing, and normalization. MLP-BranchNet is designed with multiple parallel branches that independently learn feature abstractions and are later combined to form a unified representation, enabling the model to capture complex, non-linear patterns in student data. The model achieved a test accuracy of 68.5% and a loss of 0.6323, demonstrating its effectiveness in real-world scenarios. Evaluation metrics such as precision, recall, and F1-score highlight its strengths and limitations, particularly in identifying dropout-prone students. To enhance model interpretability, we employed SHAP (SHapley Additive exPlanations) to understand the contribution of each feature in the model’s decision-making process, achieving a confidence level of 74.28%. Additionally, we analyzed the ROC curve and AUC score (0.73) to validate classification performance. An interactive dashboard was developed to visualize predictions, SHAP values, and risk analysis, enabling real-time use by educators and administrators. This work illustrates how hybrid deep learning architectures can support data-driven decision-making in educational settings.},
        keywords = {Dropout Prediction, Deep Learning, MLP- BranchNet, Educational Data Mining, Explainability, SHAP, ROC Curve, Dashboard},
        month = {May},
        }