Work place: Department of Software Engineering, Daffodil International University, Dhaka-1216, Bangladesh
E-mail: imranmahmud@daffodilvarsity.edu.bd
Website: https://orcid.org/0000-0003-2962-8515
Research Interests:
Biography
Imran F. Mahmud is currently working as Head and Professor at Department of Software Engineering in Daffodil International University, Bangladesh. He is also working as visiting professor at Graduate School of Business in Universiti Sains Malaysia. Previously, Dr. Imran worked as a senior lecturer at Graduate School of Business in Universiti Sains Malaysia. Dr. Imran was also a visiting lecturer at Institute Technology, Bandung (Indonesia) and Hong Kong Management Association (Hong Kong). He completed his PhD in Technology Management from Universiti Sains Malaysia and master‘s in software engineering from University of Hertfordshire, UK. Dr. Imran achieved several awards including ―Hall of Fame‖ and ―Presitigious Publication Award‖ from Universiti Sains Malaysia, ―Young Researcher‖ from Kasetsart University, Thailand and ―Young Scientist in Technology Management‖ from Venus International Foundation, India.
By Abu Kowshir Bitto Md. Hasan Imam Bijoy Aka Das Jannatul Ferdousi Afsana Begum Imran Mahmud
DOI: https://doi.org/10.5815/ijmecs.2025.04.06, Pub. Date: 8 Aug. 2025
This paper introduces an intelligent tool with a novel CatML stacking classifier designed to enhance predictive analytics for postgraduate university admission chances. The proposed classifier uses the CatBoost algorithm as a core component of the stacking ensemble method, which integrates CatBoost and Multi-Layer Perceptron (MLP) learners to improve predictive performance. The dataset comprises 13 questionnaire-based surveys, including academic records, standardized test scores (i.e., GRE, IELTS/TOEFL), publication status, extracurricular activities, recommendation letters, and personal statements from Bangladeshi students who applied to various U.S. postgraduate programs. Experimental results demonstrate that the CatML stacking classifier outperforms conventional models, achieving superior accuracy (88.14%) and robustness in predicting admission outcomes. The enhanced performance is attributed to the model’s ability to capture complex, non-linear relationships within the data, facilitated by the CatBoost algorithm's handling of categorical features and prevention of overfitting. Finally, this model deploys in a web system developed with HTML, CSS, JavaScript and Flask. This research underscores the efficacy of advanced ensemble techniques in educational data mining and provides a valuable intelligent tool for students aiming to navigate the complexities of U.S. postgraduate admissions. The CatML stacking classifier offers significant improvements in predictive analytics, thereby assisting students in making informed application decisions.
[...] Read more.Subscribe to receive issue release notifications and newsletters from MECS Press journals