Neural network model for detection of result anomalies in higher education

Abstract

The performance of students in tertiary institutions within and outside Nigeria is based entirely on end of academic-session examination. The processes involved in carrying out semester examinations are complex and crucial in tertiary institutions and confidentiality must be ensured. Anomaly detection in result computation is an academic problem that has been wellstudied within diverse research areas and application domains. The admission of students into different departments of tertiary institutions in Nigeria is increasing at a very high rate and has now reached a position where it is becoming difficult for the available manpower and the existing system to cope with the magnitude of Continue Assessments (CA) and exam irregularities in result computation. This leads to delay in approving students' semester results for decision making. In this paper, an efficient neural network model is developed to systematically detect anomalies in students' results as an effective measure which can enhance the efficiency and accuracy of student results. The designs of the was carried out using the Object Oriented and Design Methodology (OODM), simulated using MATLAB in the design, training and testing in order to detect result anomalies from the dataset. The model was successfully trained and tested with 96% level of accuracy with the proposed system dataset.

Keywords: Neural network, anomaly detection, academic results, object-oriented design, simulation