Predicting the Master’s Educational Achievement using Different Models

The purposes of this research were to create a model to predict educational achievement and to compare the model's efficiency by using the information of 1,468 students in the information technology graduate programs. The researcher divided the study into 2 parts: 1) to study the efficiency of the factors used as predictors and 2) to compare the algorithm performance used in prediction through unadjusted hyperparameter predictive models: Naïve Bayes (NB), Multilayer Perceptron (MLP), and Support Vector Machine (SVM) and a hyperparameter-optimized model. The results showed that the model in which the standard factor was combined with the grade point average (GPA) of the required subjects and the study plan chosen by the student had a higher predictive efficiency than a model that used standard factors alone. The performance of the algorithm without hyperparameters tuning compared to a hyperparameter-optimized model was not different. Naïve Bayes algorithm predicted by standard factors alone was able to predict the result correctly at 64.03 % whereas the model predicted by standard factors together with the cumulative GPA of compulsory subjects and the study plans was able to predict the result correctly at 82.42 %. The results showed that the GPA of required subjects and the study plans were factors that influenced the effectiveness of the predictive model.