Multiclass Grade Prediction in Higher Education via Ensemble Learning Models and Grey Wolf Feature Optimization

A challenging aspect of the academic performance prediction model over the years is to achieve high predictive accuracy and an effective feature selection approach, which traditional prediction models struggle to achieve. In this study, a new model is proposed for the prediction of student academic performance, called the Novel Scholar Grade Prediction Model (NGPM), based on multiclass classification machine learning algorithms. The aim of this study is to correctly predict the final grades of students on the input data by identifying and laying out suitable features. For model development, a dataset of student academic records from a Higher Education Institution in Central India was used. In this novel model, the selection of feature was addressed, which is optimized using the Grey Wolf optimization (GWO). This model combines the SVM, K-NN and LSTM Networks with the bagging ensemble method to create a strong and precise prediction model. In the academic session, the evaluation results showed an average precision value of 0.9975 and recall value of 0.9920 on the data of students achieving Grade A at the 40% course completion with an overall accuracy improvement of 15% as compared to the earlier models applied. The ensemble model gives a significantly better predictive accuracy when feature selection is tuned using GWO.