A Comparative Study of Hybrid Neural Network with Metaheuristics for Student Performance Classification

Abstract

This study investigated the use of an neural network (NN) and metaheuristic algorithms for predictions using the Higher Education Students Performance Evaluation Dataset. An NN with the backpropagation (BP) algorithm is a widely accepted machine learning method that uses past data for prediction and classification, while metaheuristic algorithms can be used to find better subsets of input variables to import into the NN, hence enabling more accurate predictions by reducing errors. The imported data were chosen from the UCI Machine Learning Repository. MSE and MAE used to evaluate the hybrid intelligence approach with respect to reducing prediction errors. The experimental results showed that optimal efficiency was obtained using an NN with the student psychology-based optimization (SPBO) model, while competitive metaheuristic algorithms.