Predicting Student Final Score Using Deep Learning

Abstract

The purpose of this paper is to create a smart and effective tool for evaluating students in classroom objectively by overcoming human subjectivity resulting from lack of experience of instructors, and students’ over-trust in themselves. We had provided instructors in Sur university with the “Program for Student Assessment (PISA)” tool to assess it’s positive impact on academic performance, self-regulation, and improvement on their final exam scores. The study sample included in this study was the students enrolled at Sur University College at the time of data collection in the 2018/2019 semester. In the purpose of testing the efficiency of four models in predicting students’ final scores based on their mark in the first exam. The four tested algorithms were: Multiple Linear Regressions (MLP), K-mean cluster, Modular feed for-ward neural network and Radial Basis Function (RBF) (De Marchi and Wendland, Appl Math Lett 99:105996, 2020 [3], Niu et al, Water 11(1):88, 2019 12]). After comparing the four models’ effectiveness in predicting the final score, results show that RBF has the highest average classification rate, followed by neural network and K-mean cluster, while Multiple Linear Regressions was the worst at performance. RBF has been used to create the Instructor Program for Student Assessment (PISA).predicting student performance early will help students to improve their performance and help instructors modify their teaching style to fit their student’s needs.