Abstract:
This study explores the possibility to improve students' math proficiency prediction with a multi-class neural network model by augmenting the training dataset with synth...Show MoreMetadata
Abstract:
This study explores the possibility to improve students' math proficiency prediction with a multi-class neural network model by augmenting the training dataset with synthetic data. The original training dataset was based on the publicly released PISA 2012 computer-based database. Three math proficiency classes were established: low, mediocre and high. Minority class with the least number of samples was the high proficiency class. SMOTE, VAE and CTGAN methods were used to augment the minority class with additional data samples. G-mean was used as a performance measure for observing the enhancement to the prediction of the minority class.
Published in: 2022 IEEE 20th Jubilee International Symposium on Intelligent Systems and Informatics (SISY)
Date of Conference: 15-17 September 2022
Date Added to IEEE Xplore: 13 February 2023
ISBN Information: