Journal of Engineering Education Transformations

Abstract

Abstract : Researcher interest in education data mining has increased significantly in a variety of sectors. The recent research works use a variety of machine learning techniques to predict students' academic success in the educational sectors. They suffer from serious drawbacks like low forecast accuracy, high processing times, and overhead. Therefore, the proposed work aims to develop a new model for projecting students' academic progress. The main goal of this paper is to develop a smart and automated system for predicting the students’ academic performance from the given students’ data. For this purpose, a novel optimization and deep l earning classification methodologies are implemented in this study. Here, the public UCI education training dataset is obtained to develop the prediction framework for forecasting students' academic achievement. The most correlated features from the preprocessed schooling dataset are chosen using the War Strategy Optimization (WStO) method to improve predicting performance. To effectively and reliably estimate the student performance rate with few wrong predictions, a classification method based on the Bi-directional Gated Recurrent Neural Network (Bi-GRNNet) is applied. The Arithmetic Operation Optimization Algorithm (AO2A) is used to correctly optimize the parameters of deep learning classifiers to guarantee minimal computing system complexity and quicker training. By using a complete performance evaluation study that takes into consideration a variety of various parameters, the output of the proposed WStO + Bi-GRNNet model is validated and analyzed. According to the findings, it is inferred that the proposed Bi-GRNNet integrated with WStO and AO2A technique performs well and provides an increased accuracy up to 99% while effectively predicting the students’ academic achievements.

Keywords: Education Data mining, Students’ Performance Prediction, Academics, War Strategy Optimization (WStO), Bi-directional Gated Re curr ent Neur al Network (Bi -GRNNet ) Cl ass i fica tion, and Ari thmet ic Operat ion Optimization Algorithm (AO2A).