
Mehedi Hasan
Lecturer
At a glance
Biography
Mehedi Hasan is a technically intuitive aspiring practitioner with deep interests in Machine Learning, Image Processing, Artificial Intelligence, Deep Learning, Computer Vision, Network Security, and Cybersecurity. His academic excellence has been recognized through two consecutive Dean's Awards in 2023 and 2024, reflecting his strong analytical aptitude and dedication to computational innovation. He actively strengthens his skills through self-directed projects, rigorous coursework, and exploration of intelligent system design. With a sharp problem-solving mindset and commitment to technical mastery, Mehedi aims to develop impactful AI-driven solutions for complex real-world challenges.
Education
| Degree | Institution | Year | Result |
|---|---|---|---|
| B.Sc. in Computer Science and Engineering | Dhaka University of Engineering & Technology (DUET) | 2024 | |
| Diploma in Engineering | Rangpur Polytechnic Institute, Rangpur | 2019 | |
| Secondary School Certificate | Gajaghanta High School and College | 2015 |
Research Interests
Publications
Conference Publications
Predicting Student Dropout and Academic Success Using Explainable Ensemble Machine Learning
Student attrition in higher education has a negative effect on educational institutions and on student career success. While machine learning models have demonstrated their ability to predict, they are not necessarily easy to interpret, which can make it difficult to adopt them in the field of education. In this work, we propose an interpretable stacking ensemble model to predict student dropout and academic achievement from a public dataset with 4424 student records. To capture the performance trend, six temporal academic features were engineered, and class imbalance was solved by a leakage-free SMOTE method, applied only to the training data. The model proposed will use six base classifiers (XGBoost, Random Forest, Extra Trees, Gradient Boosting, LightGBM, CatBoost) with a Logistic Regression meta learner. The results of the experiments reveal that the stacking ensemble performed the best with an overall accuracy of 81.02% and a weighted F1-score of 80.26% for predicting Graduate, Dropout, and Enrolled outcomes. To increase transparency, SHAP and LIME were used to explain both global and local explanations of model predictions, showing that factors such as student approval rates, grade progression, and cumulative academic achievement are among the most impactful and associated with student outcomes. The proposed framework makes accurate and interpretable predictions that enable data informed decisions of student retention and interventions.
Academic Experience
Lecturer
22 January,2025 - PresentDepartment of CSE at Bangladesh Army University of Engineering & Technology
Honors & Awards
Two consecutive Dean’s Award in 2023 and 2024.
Two consecutive Dean’s Award in 2023 and 2024.
Contact
Send a message to Mehedi Hasan through the department office. Direct email and phone details are listed above.