Dropout Prediction of University Students in Bangladesh using Machine Learning Technique

Nowadays, among Bangladesh's educational institutions, the number of university students dropping out is high. A high dropout rate is a problem that presents many negative consequences. It is an academic issue for university students, is a significant risk to local communities and countries and also carries an unfavorable economic impact. Dropout occurrences typically stem from either academic or personal factors. This study aims to find out the factor behind the high rate of semester dropout in Bangladesh as well as predict the precarious zone for dropout so that the massive rate of dropout can be stopped and to develop methodologies for detecting individuals who may be inclined to stop their academic pursuits. For effective dropout prediction, 301 data have gathered by surveying data by using google survey form keeping 25 questions from various universities in Bangladesh. For effective dropout prediction, several preprocessing techniques have been applied on collected data as well as KNN, Logistic Regression, Random Forest, XGBoost, LightGBM, AdaBoost, Bagging and SVM were applied on the preprocessed data. Among the models SVM and LightGBM performed better than all other working classifiers attaining 96% and 95% accuracy. Application of these studies supports University authorities in identifying early dropouts, refining academic policies to prevent dropouts and also enhancing student engagement, improving long-term institutional planning and resource optimization in universities.