A Deep Learning Approach to Recommending Undergraduate Programs for Bangladeshi Students

Academic success and satisfaction are strongly influenced by the alignment between students’ personality traits, cognitive abilities, and their chosen academic programs. This study investigates how these factors vary across students in nine academic disciplines in Bangladesh and proposes a deep learning-based recommendation system to guide students toward programs that best match their profiles. Data were collected from 233 participants, including demographic information, academic performance, personality traits (assessed using the Big-5 model), and reasoning ability (measured via Factor B of the 16PF assessment). A deep learning model incorporating three dense layers, ReLU activation functions, and dropout layers was developed to prevent overfitting. The model achieved an accuracy of 94.5%, outperforming traditional methods such as Decision Tree and Random Forest classifiers. These findings demonstrate the potential of deep learning frameworks to enhance academic decision-making, fostering greater student success and personal growth.