MACHINE LEARNING-BASED RNN-LSTM FOR PREDICTING COVID-19 CASES IN MALAYSIA

Abstract

Listen

Translate article

Started in the beginning of January 2020, the world still struggles to cope with the spread of COVID-19, including in Malaysia. Although the world has invented some vaccines for the disease, the coronavirus problem is still there and perhaps it will be a big issue until a few years ahead. This study will look at the trend of the spread on COVID-19 in Malaysia after two years since the pandemic haunted human life, particularly when the pandemic turned into endemic. We developed a model prediction based on Recurrent Neural Networks (RNN) with Long-Short Term Memory (LSTM) to predict the COVID-19 outbreak in both pandemic and endemic periods. In addition, the COVID-19 classes are split into susceptible (S), exposed (E), infectious (I), and recovered(R). We then forecast 60 days ahead by using these two models which are RNN with long-short term memory (LSTM) and a support vector regression (SVR). The results prove that both methods have different advantages. SVR can perform better in predicting the pandemic period, while RNN-LSTM has better predict the endemic period. From the results, it can be said that SVR is more appropriate for predicting dynamic curves, while RNN-LSTM is suitable for smooth curves. Thus, the results show that the studied prediction technique is a promising and encouraging procedure for studying the future trends of the COVID-19.