ABSTRACT
Universities play a notable role in the development of a country by producing skilled graduates for the country. The primary reason for non-degree completion is academic failure. This research models “An Artificial Neural Network” and uses it to predict the class of degree a university student will graduate with, at the end of the completion of his/her programme of study. Population of study consisted of students in their first or second year in the 2021/2022 and 2022/2023 session who enrolled for a Bachelor degree programme at University of Benin. A centralized repository of data was built for the neural network from data collected from 141 graduates. Samples were selected from this data and used for testing the predictive accuracy of the model. By using stratified sampling a sample of 100 students was selected, questionnaires was administered to them to elicit demographic, academic and socioeconomic data from them and their final class of degree was predicted and they were classified as ‘at risk’ or not ‘Not at risk’ with respect to their class of degree at the end of their study time. SSCE grades, Post UTME scores, Age, Gender, Study habits, class attendance and socioeconomic status were used as input to the Neural Network., The output was the students’ predicted class of degree, which was either First Class Honors, Second Class Honors (Upper Division), Second Class Honors (Lower Division), Third Class Honors or Pass. In classifying the students at risk/not at risk, an accuracy rate of 96% was achieved. Findings from the study revealed that an artificial neural network can accurately predict students’ academic performance to a great extent. It was also established that the integration of certain demographic factors like age and gender, socioeconomic status, study habits, class attendance and previous academic scores and grades are pivotal in modelling, predicting and classifying students’ academic performance. It was recommended that a holistic student assessment should be practiced, there should be an enhanced data integration in predictive modelling, educational researchers and practitioners should embark on a collective pursuit to enhance the transparency and interpretability of neural networks and a collaborative research network should be embraced.
