Prediction of Students' Academic Performance using Artificial Neural Network

Universities play a remarkable role in the development of a country by producing skilled graduates for the country. Graduation rate is low as compared to the enrollment rate in the higher education institutions. Academic failure is main reason for non-degree completion. Students' retention and high academic performance are significant for students, academic and administrative staff of universities. In this paper, our objective is to Predict the chance of students being at risk (AR) or not 'Not at risk' (NAR) with respect to their degree. Population of study consisted of all students of social sciences studying in 4th semester and they enrolled in 2007 session of BS and MA/MSc program at University of Gujrat Hafiz Hayat Campus. By using stratified sampling with proportional allocation method, a sample of 300 students was selected. We have used Multilayer Perception Neural Network Model to predict the chance of students being at risk (AR) or not 'Not at risk' (NAR) with respect to their degree on the basis of CGPA at the end of 2nd semester, Study time. Previous degree marks, Home environment, Study habits Learning skills, Hardworking and Academic interaction. In classifying the students at risk/not at risk, we could achieve a rate of correct classification of over 95% in training sample and over 85% in holdout sample. The estimated models can be used to predict the students being at risk or not with respect to their degree.