A novel approach to mitigate academic underachievement in higher education: Feature selection, classifier performance, and interpretability in predicting student performance

Abstract

The main goal of this study is to address the ongoing problem of low academic performance in higher education by using machine learning techniques. We use a dataset from a higher education institution that includes various information available at student enrollment, such as academic history, demographics, and socio-economic factors. To address this issue, we introduce a new method that combines the Slime Mould Algorithm (SMA) for efficient feature selection with a Forest-Optimized Neural Network (FO-NN) Classifier. Our method aims to identify students at risk of academic failure early. Using the SMA, we simplify the feature selection process, identifying important attributes for accurate predictions. The Forest Optimization technique improves the classification process by optimizing the neural network model. The experimental results of this study show that our proposed method is effective, with significant improvements in feature selection accuracy and notable enhancements in the predictive performance of the neural network classifier. By selecting a subset of relevant features, our approach deals with high-dimensional datasets and greatly improves the quality and interpretability of predictive models. The innovative combination of the SMA and the FO-NN classifier increases accuracy, interpretability, and the ability to generalize in predicting student performance. This work contributes to a more effective strategy for reducing academic underachievement in higher education.