A Dual Attention-Driven Neural Ensemble Framework for Student Outcome Prediction

Education cultivates individuals' cognitive, social, and interpersonal characteristics and is the basis of both individual and society development. Higher secondary school level Education is especially important as it integrates classroom teaching to deeper academic or professional aspirations. This is an important phase in a student's educational journey as future possibilities tend to be determined by the standards of education received at this stage. Therefore, attaining success at this level has significance for both personal development and the development of society. Initial academic achievement prediction has emerged as an acceptable method for assisting students who are at risk of performing poorly in such an environment. The methodologies of Machine learning (ML) and deep learning (DL) have been used for determining a range of educational, demographic, and socioeconomic factors in aiming to more precisely predict student outcomes. The features of these prediction models help educators and lawmakers to develop targeted interventions, enhance instructional techniques, and allocate resources. This study evaluated the extent to which early higher secondary educational achievement is predicted by ensemble learning approaches. The proposed ensemble method combines two deep learning methodologies, i.e., MLP (Multilayer Perceptron) and DANet (Dual Attention Network). The approach was validated using an OULAD dataset of approx. 38000 students. The experimental outcomes indicate that the proposed ensemble model surpasses MLP with 95.86 % and DANet with 95.72 %, with a maximum overall accuracy of 96.20%.