Prediction of Student Performance Using Rough Set Theory And Backpropagation Neural Networks

Abstract

With the rise of web-based education systems and the increased use of information systems in education institutions, the amount of data recorded on student performance and behavior has increased exponentially. Thus, bringing about a large number of contributions to the field of educational research, which in itself contributed to the further evolution off the field in the last two decades alone, with terms such as Educational Data Mining (EDM), Learning Analytics, Data-driven Education, Teaching Analytics and others being added to the literature. In this paper, we evaluate the usefulness of a model using Rough Set Theory (RST) and Backpropagation Neural Network (BPNN) in effectively predicting the students’ overall performance. The dataset used consists of 10 different attributes and one decision factor belonging to 53 students collected from a language course which administers in-person education with the aid of an online platform for assignments. RST was implemented in order to reduce the number of attributes used as input in the neural network and the BPNN made an accurate prediction using only 5 of the initial attributes. Thus outperforming a model based solely on BPNN used on the original dataset and reducing computational costs.

Enhancing Student Performance Prediction through LSTM-based Deep Learning Models with Unbalanced Data Handling using Oversampling Approach

“…According to the research done [15], the MLP classifier with the SMOTE technique performs better than the ML algorithm used. A significant improvement in accuracy can be achieved by basing the model's development on the input of particular variables [8], [10], [18], [20]. LSTM is a subtype of recurrent neural networks (RNN) [33].…”