Smart Retention: AI-Driven Approaches for Predicting Student Dropouts in Higher Education in Bangladesh

Student dropout refers to students leaving their academic pursuits prematurely, which negatively impacts educational outcomes and resource utilization. In Bangladesh, this challenge is particularly vital due to the rapid expansion of access to higher education, leading to high dropout rates that hinder individual progress and the overall quality of education. To address this issue, we apply machine learning and deep learning techniques to predict and mitigate student dropouts in Bangladeshi higher education institutions. These models analyze diverse student data, including demographics, educational, and socioeconomic factors. We generated a data set through an online survey, collecting 3119 instances, which was then preprocessed using techniques such as handling null values, removing duplicate values, label encoding, and over sampling. The dataset was split into 2414 training samples and 603 test samples. Various algorithms were utilized for this task, including Decision Tree, XGBoost, Logistic Regression, K-Nearest Neighbors (KNN), Bernoulli Naive Bayes, LightGBM, Artificial Neural Networks (ANNs), and Bidirectional Long Short-Term Memory (BiLSTM). The models achieved impressive results, after training and evaluation, with Random Forest achieving an accuracy of 97.35% and a deep learning model, CNN, achieving 95.2% accuracy. These results show the potential of ML and DL in predicting student dropout and offer insights to further enhance student retention.