Toward Prediction of Student’s Guardian in the Secondary Schools for the Real Time

Abstract

To ease of school management for the identification of student’s protector during his or her schooling and evaluate the performance of a student, the guardian predictive models are presented with the help of machine learning algorithms. For this, standard secondary dataset of two secondary schools was considered from Portugal belonging to the language course. The initial dataset consisted of 649 instances and 33 features. These features belong to the student’s academic, demography, family and personal features. The guardian feature has been considered as a class variable and others (significant) features assumed as predictors. In the orange platform, three machine learning algorithms, support vector machine (SVM), random forest (RF) and neural network (NN), were used with three testing techniques. On one hand, the SVM computed the highest prediction probabilities of 0.996 for other class and another hand, the NN gives the largest prediction probabilities such as 0.906 for father class and 0.889 for mother class. The NN attained the most guardian prediction accuracy of 89% and outperformed others. Also, leave-one-out method significantly enhanced the prediction accuracy of each learner except the SVM. Also, it proved the NN learner slower with prediction time (23 s) and makes the SVM as faster with time (14 s). This study may not only helpful to the school management but also support the social administration of the district or state. Using the model, it must be significant to predict the care-taker of the student.