Predicting Board Performance Using Classification Algorithms and Time Series Analysis

Higher Education Institutions are increasingly recognizing the value of Educational Data Mining (EDM), a method that collects and analyzes educational data. EDM can employ machine learning techniques to assess and predict students’ academic outcomes. This paper aims to apply EDM, integrating time series analysis and classification algorithms, to assist a university in the Philippines in improving its students’ performance on the Licensure Examination for Teachers. The study involved an in-depth literature reviews and applied feature selection techniques to determine the significance of specific attributes. Subsequently, a comparative analysis was undertaken among three classification algorithms to ascertain the most accurate model in predicting student’s examination performance. Various validation metrics were used to validate the results. Finally, the most optimal model was selected and integrated into the system. Moreover, the Rapid Application Development (RAD) model was employed to develop the “Board Prophet” system, which offers various functionalities. These includes an online dashboard for accessing historical data, time series analysis capabilities for predicting the university’s board performance, individual predictions of a students’ success or failure in the licensure examination, and proposed intervention plans based on the system’s predictions. Furthermore, the study’s findings revealed that the student’s General Weighted Average and Review Center were the two most significant attributes, with Decision tree outperforming other classification algorithms. The system features were assessed using ISO/IEC 25010 standards and the Technology Acceptance Model (TAM), with users showing high acceptance. These findings will empower the university make better decisions that will improve their students’ licensure examination performance.