Towards Prediction of Academic School of Students for Real-Time

This study presented the five predictive models to predict the student's study school based on their demographics, personal, social, and grade features. Using the Orange stable version 3.23.0 machine learning tool, the authors performed numerous experiments to identify the study place (school) of students. The dataset stored the records and attributes of two senior secondary schools in Portugal. A binary classification problem has solved using the five machine learning algorithms. Also, the training time and testing time of each classifier have been compared. After the preprocessing, 875 instances and 31 attributes were practically analyzed in the Orange workflow environment. This experiential study's findings presented five predicted models and proved that Artificial Neural Network outperformed others with providing the most accuracy of 84.7%. The second winner classifier Random Forest attained an accuracy of 83.7% with 10-folds. Increasing holdout ratios and folds significantly improved the prediction accuracies of each classifier. Also, full cross-validation (leave one out) is recommended for this dataset compared to others. Based on the testing time of 8.2 seconds, the Random Forest classifier has been suggested, and the Neural Network is recommended owing to the highest prediction accuracy. The school predictive models supported identifying the school to be aware of student situations in their academic, personal, and social life.