FUTURE PREDICTION OF COVID-19 IN INDONESIA USING DEEP LEARNING

Abstract

Lack in machine learning methods are the inability to predict more than one step forward, because in each step forward machine learning must calculate the error value with real data to be used to correct the weight value in the next round. This study aims to predict the long-term future of Covid-19 in Indonesia. The method used is to make the results of one-step predictions from machine learning as the first new data. Then in the next iteration machine learning will produce one more prediction step as the second new data, so we get two prediction results going forward. This process is repeated until it reaches the desired long-term prediction of 2 months (50 days). Long prediction for the next 2 months, is done by using 2 Deep Learning (DL) methods, namely Long Short Term Memory (LSTM) and Gated Reccurent Unit (GRU) in the 4 hidden layers of learning machine used. The GRU model on the four hidden layers gives the best results with a value of RMSE = 206,632 at epoch = 5.