Students’ Performance Prediction using Artificial Neural Network

Abstract

The increasing availability of educational data has created opportunities for applying advanced machine learning techniques to improve student outcomes. This study focuses on predicting students’ academic performance using Artificial Neural Networks (ANN). The aim is to develop a predictive model that accurately estimates student performance based on academic, behavioural, and socio-demographic factors, such as attendance, IQ, study habits, and prior academic records. The data for the study contained both primary and secondary data to ensure data triangulation. The primary data collected through a questionnaire was obtained from 720 respondents from Adamawa State Polytechnic, Yola. Secondary data containing mainly students' cumulative grade point average (CGPA) was obtained from the students' academic record division of the school management information system (MIS) office. The structured dataset collected was pre-processed to ensure data quality, including normalization and handling of missing values. The ANN model was designed with input, hidden, and output layers, and trained using backpropagation to learn complex relationships within the data. The model’s performance was evaluated using a confusion matrix. The prediction accuracy of 97.36% and the standard deviation of 0.52% was obtained. The results indicate that the ANN model achieved a high level of prediction accuracy, outperforming traditional statistical approaches in capturing non-linear patterns among variables. This study demonstrates that Artificial Neural Networks can serve as an effective tool for early identification of students at risk of poor academic performance. The findings provide valuable insights for educators and institutions seeking to implement data-driven strategies for improving student success. Future work is recommended to enhance model transparency and incorporate hybrid approaches for improved predictive performance.