Survey of Predictive Analytics Models for Student Dropout in Higher Secondary and Higher Education

Dropout in schools remains a major issue in upper secondary education institutions and higher education institutions in their current form because it impedes educational processes and results in long-term socio-economic implications. Recent research in Artificial Intelligence (AI), Educational Data Mining (EDM), and Machine Learning (ML) has facilitated more precise early prediction of dropout individuals by identifying complex patterns present in behavioral traces, academic records, demographic variables, and Learning Management System interactions. However, in current studies, researchers often resort to Random Forests, Logistic Regression Analysis, Support Vector Machines (SVM), Decision Trees, Ensemble Learning approaches, and Deep Learning methods that often seek improved predictive accuracy. This study reviews ongoing research trends, approaches, and predictive variables in the field of dropout prediction research and asserts the importance of data preprocessing steps, feature selection variables, data balancing techniques, and transparency parameters in designing predictive models that are studied in a reliable framework. This research proposes that dropout rate at the higher secondary school level remains a challenge for educational continuity. This research uses models based on machine learning created using the Scikit-learn framework for the prediction of dropout students based on academic-related parameters. Various models have been used for class predictions, using which patterns related to early school dropout can be identified. The study demonstrates that models developed using Scikit-learn are useful for academic early intervention.