Enrollment confirmation prediction for prospective university students by applying deep learning techniques

Abstract

The objectives of this research are 1) to study the models of enrollment confirmation prediction for prospective university students, 2) to study and find the appropriate parameter values to enhance the efficiency of the models, 3) to compare the model performance of deep learning techniques and traditional machine learning techniques for predicting the university students' admission confirmation. The applied models include Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Random Forest, Naive Bayes, and Support Vector Machine (SVM). The samples are 4,479 records of students who applied for admission to universities during the academic year 2018-2021. The modeling of traditional machine learning and deep learning uses 11 features consisting of prefixes, schools, applicant types, and 8-subjects grades. The performance comparison between the two models revealed that the deep learning model outperformed the traditional machine learning model. Therefore, a study was conducted to find suitable parameters to enhance the efficiency of the deep learning model. The results of the study found that the CNN model had the highest accuracy at 64.68 percent, and the LSTM needed to be more suitable due to overfitting. In conclusion, the CNN was suitable for applying to the system of enrollment confirmation prediction for prospective university students.