Factors of Pre-University Study in Influencing Graduate on Time

In education, identifying the important pre-university study factors poses a formidable challenge, due to the necessity for human intervention in the variable selection process. Determining the important factors enables the educational institutions to shape strategic planning throughout a student’s entire academic journey, with a specific focus on the pre-university factors. Researchers have investigated various approaches to identify the optimal features. One of the techniques is by leveraging Artificial Intelligence. Therefore, this study has attempted to (i) identify the important pre-university study variables towards graduate on time, and (ii) develop as well as evaluate predictive models with and without the important pre-university study variables identified. The dataset is collected from a university in Malaysia, containing 4463 student records and 27 variables. It consists of student profile, registration details, grades of Sijil Pelajaran Malaysia, and English Test. Data preprocessing was performed to remove those features with more than 75% missing values and impute other features with median. Important features was identified by applying Boruta, and BorutaShap. Among 22 variables, both methods identified permanent address district, date of birth, campus, program description, and the grades of History subject. To evaluate the influence of the important factors towards graduate on time, several Machine Learning models were constructed due to its simplicity and evaluated with and without the important features identified based on accuracy, precision, recall, F1-score, and Area under the Curve, The findings showed that Gaussian NB produced the highest recall (95.2%) when the important features identified by BorutaShap were used.