Bidirectional Long Short-Term Memory Integrated Text-to-Text Transfer Transformer to Identify Predictive Variables for Students’ Academic Performance Analysis

Abstract

Educational institutions and professionals must analyze student performance to identify ways to improve each student's academic achievement. Predicting how well students will do in universities and colleges is crucial for managing their time effectively. Predicting students' academic achievement has been the subject of some relevant research activity since the advent of deep learning. Students' academic performance is affected by various factors, yet there is a general trend of similar intelligence and learning styles among students at the same university. The goal of this research is to use deep learning algorithm and historical data to forecast how well students will do in academics. Academic performance analysis can be enhanced by combining a Text-to-Text Transfer Transformer (T5) with a Bidirectional Long Short-Term Memory (Bi-LSTM) model. This allows us to leverage the capabilities of both architectures. In contrast to Bi-LSTM's adeptness at capturing sequential dependencies, T5's text-to-text method offers a versatile framework for encoding and processing a wide range of academic data. Accuracy in activities such as requirements classification and short response grading can be enhanced using this hybrid technique. The proposed T5-BiLSTM model is used to accurately identify the predictive variables from academic datasets that majorly influence the academic performance. This research is helpful for educational institutions to identify the predictive variables that are used to enhance the student academic progress.