Abstract
Student dropout in higher education remains a significant challenge, particularly in rural and semi-urban regions where academic, financial, and institutional factors influence student retention. This study investigates the determinants of undergraduate student dropout in Maihar District, Madhya Pradesh, and develops a machine learning-based framework for early dropout prediction. Primary data were collected from six undergraduate colleges (three government and three private) through structured questionnaires and institutional records.
The dataset comprises 315 student instances with variables including regular absence from college, academic performance, family income, parental education, transportation accessibility, digital connectivity, psychological well-being, lack of educational facility, without adult protection and institutional support. After data preprocessing and feature selection, three supervised machine learning algorithms-Logistic Regression, Decision Tree (J48), and Random Forest- were implemented using the WEKA data mining tool. Model performance was evaluated using Accuracy, Precision, Recall, F1-Score, ROC area, and confusion matrix. Among the models, Random Forest achieved the highest classification accuracy of 93.02%, followed by J48 (88.57%) and Logistic Regression (84.80%).
The results indicate that regular absence, academic performance, family income, digital connectivity, and weak institutional support are the most influential predictors of student dropout. The proposed framework serves as an effective early warning system, enabling timely interventions to improve student retention and support sustainable higher education development.
KEYWORDS
Student Dropout, Machine Learning, Educational Data Mining, Random Forest, Predictive Analytics, Logistic Regression, Student Retention.
Shivendra Kumar Dwivedi
Department of Computer Science, Sharda Mahavidyalaya, Sarlanagar Maihar-485772, M.P., India