Graduates' Prediction System Using Artificial Intelligence

Graduation rates are an essential metric for universities to gauge the effectiveness of their programmes. Prediction models can be used to assess students' likelihood of completing their degrees, as well as to analyse the rate of completion of programmes that they offer. Several studies have been done previously using predictive models in various universities around the world. At King Mongkut's University of Technology North Bangkok (KMUTNB), Prachuabsupakij and Wuttikamolchai developed a web application that used the Decision Tree Algorithm to predict student's graduation. Other studies made use of other machine learning algorithms such as support vector machine, neural network and classification and regression tree algorithms. These studies achieved a successful prediction rate of an average of above 70% accuracy. The University of Namibia does not have any predictive models of any kind and neither has any study of this nature ever been done. It was for this reason that this study was conducted using the School of Computing as the case study. The study would then develop a graduates' prediction system for the school of computing. The system is a web-based application that requires a user to log in before accessing its features, such as displaying the predicted outcomes. A sample of 500 student data was used to create the student dataset. The prediction was done using four different prediction models which were then comparatively analysed. Those prediction models are Neural Network, Decision Tree, Support Vector Machine and Random forest. The web-based application consists of an interactive dashboard to allow the user to visualize the prediction results which can be view using different types of charts all to the user's convenience. This study managed to attain its main object of creating a student graduation prediction system for the school of computing and the sub-objective were also met. The predictions of this study were purely based on the students' academic results in the form of credit scores and other factors were not considered. This study recommends that further research should be done with actual student data from the university's student database records that stretches for over a longer period, say about 5 to 10 years.