Students’ Performance Prediction in Higher Education During COVID-19 Pandemic Based on Recurrent Forecasting and Singular Spectrum Analysis

Abstract

The COVID-19 pandemic is a virus that is changing habits in human life worldwide. The COVID-19 outbreaks in Indonesia have forced educational activities such as teaching and learning to be conducted online. Teaching and learning activities using the online method are familiar, but the effectiveness of this method still needs to be investigated to be applied in all educational systems. This study used the predictive modeling of Recurrent Forecasting (RF) derived from Singular Spectrum Analysis (SSA) to know the online learning method's practicality on the student's academic performance. The fundamental notion of the predictive fusion model is to improve the effectiveness of several forms of forecast models in SSA by employing a fusion method of two parameters, a window length (L), and a number of leading components (r). This study used undergraduate students' grade point averages (GPA) from a public university in Indonesia through online classes during the COVID-19 epidemic. The experiments unveiled that a parameter of L = 14 ( ) yielded the finest prediction using the RF-SSA model with a root mean square error (RMSE) value of 0.20. Such a finding signified the ability of the RF-SSA to project the students' academic performance according to the GPA for the forthcoming semester. Nonetheless, developing the RF-SSA algorithm for greater effectiveness is essential to acquiring more datasets, such as by gathering a bigger group of respondents from several Indonesian universities.