|
|
2020 » Papers » Volume 2 » NATION-WISE AFFILIATION PREDICTION FOR THE REAL-TIME 1. NATION-WISE AFFILIATION PREDICTION FOR THE REAL-TIME Authors: Verma Chaman, Stoffova Veronika, Illes Zoltan Volume 2 | DOI: 10.12753/2066-026X-20-121 | Pages: 284-290 | Download PDF | Abstract
To support the real-time prediction of the student demographic features such as gender, age, study level etc., in different type of online surveys and questionnaires machine learning played a vital role. In this paper, the author's applied a five supervised machine learning algorithms to predict the student's association with their home land (native) towards the technology used in their university. For this, a primary dataset has been collected with google form from the two country university (Indian and Hungarian. The present model might be useful to identify the student category based on their native nation towards their thinking for the technology in higher education. The identification of the nation affiliation might help to understand the satisfaction, usability, technology available and attitude of student for the ICT as well. The best predictive model may be deployed on real-time web module to support demographic prediction system towards ICT. Considering the binary classification problem, the student's nation wise affiliation class set to as a response variable and rest of features was considered as predictors. A student's nation wise affiliation class has two values such as Hungarian student and Indian student. In the Weka experimental environment, the primary dataset of 331 records with 38 attributes were analyzed using the Logistic Regression (LR), Support Vector Machine (SVM), Multilayer Perceptron (MP), K-nearest neighbor (KNN) and Random Forest (RF). The performances and prediction time of each classifier was compared with statistical t-test at the 0.05 confidence level. The authors performed two major experiments with training ratio and cross-validation method using these classifiers. The first experiment found no significant difference between the accuracies of the classifiers in the affiliation prediction. Also, the RF classifier outperformed others with the highest prediction accuracy of 87.56% in 0.07 seconds. The findings of the second experiment proved that the RF attained the highest prediction accuracy of 89.33% as compared to others in 0.08 seconds. This experiment also evidenced the prediction accuracies of the KNN, MP, and LR are found a significant difference as compared to the RF and SVM classifier. The statistical t-test also proved that the prediction times of the SVM, KNN, and LR found significant different than others in each experiment. Hence, the authors presented the RF model to identify the student's affiliation (Indian or Hungarian) towards technological use after made a comparison of various performance measures. | Keywords
Affiliation prediction; Machine learning; Real-time; ICT. |
|
|
|