Deep Learning Based An Optimized Predictive Academic Performance Approach

Abstract

An early warning system is used to collect, process and analyze present data to predict possibilities that may occur in the future. This tool can be implemented in education to process relevant data to predict academic performance and threats. Several studies have been conducted in the past several decades on the use of early warning systems in education. Moreover, there are limited open datasets available in these areas of research. A remarkable dataset is the Open University Learning Analytics Dataset (OULAD). This paper proposes a deep learning-based predictive analytics model with an effective specificity score that helps predict student academic performance. Moreover, the paper analyzes the implementation timing of the model within the first two months of the academic semester. The model attains a higher success prediction accuracy rate within multicategories and a large input dataset. The best significant result achieved in the study was the 98.94 accuracy score and 93.10 specificity score in the first week of Science Technology Engineering Mathematics (STEM) domain courses of the academic term, compared to Artificial Neural Network, Naive Bayes and Support vector machine, which were applied as validators.