Student performance prediction based on multiple-choice question test using neural network in the VLab platform

Abstract

Educational institutions are now concentrating to develop new initiative methods for recognizing the ability of the students through evaluation of the student’s answer sheets. The manual evaluation of the answer sheet is a burdensome process for the tutors, which consumes more time and increases stress of the tutors. Hence, the advanced method is required for the automatic assessment of the student’s mark sheet which saves the time of the tutors and meets the demand of the educational institution. In this research, the student performance is evaluated through the MCQ test using the neural networks in the VLab platform. The computational complexity and the overfitting issues are greatly reduced by the feature extraction process through Term Frequency-Inverse Document Frequency (TF-IDF) technique. The effectiveness of the proposed method is manifested through the comparative analysis. The accuracy, precision, and recall attained by the proposed student’s performance prediction based on Neural Network while considering training percentage are found to be 0.9416, 0.9364, and 0.9502 respectively. The accuracy, precision, and recall attained by the proposed student’s performance prediction based on Neural Network, while considering the K-fold value are found to be 0.9475, 0.9474, 0.9538, respectively.