Researchers have utilized a modeling study to predict the potential spread of the Covid-19 pandemic in Indonesia. The optimistic prediction results predict Covid-19 ending in the last week of May, 2020.
The model used in this prediction study is a probabilistic model based on real data or Probabilistic Data-Driven Model (PDDM).
Researchers involved in this study are Professor of Statistics at the University of Gadjah Mada (UGM), Prof. Dr. rer nat Dedi Rosadi SSi MSc, UGM MIPA alumni Drs Heribertus Joko Kristadi MSI, and PPRA alumni of the Indonesian National Defense Institute Dr. Fidelis Indriarto SSi MM.
What is the PDDM model?
The PDDM model used by researchers is a theory model that assumes the process of patients coming to the hospital as confirmed cases of Covid-19, following the Markovian queuing process.
After matching the model to the total data confirmed cases of Covid-19, the researcher was able to explain many important phenomena based on the model used.
The PDDM model is a refinement of the basic statistical model developed by Heribertus Joko Kristadi.
Dr Dedi stated that the PDDM model has been tried and compared with various statistical models, machine learning, and time series such as the Gompertz curve, Logistic and Exponential Models, ARIMA, and others.
But according to Dedi, the PDDM model is better for describing total data of patients with Covid-19 than for predictions based on dynamic mathematical models.
Dr Fidelis added that the results of bombastic analysis and inaccurate estimations previously, were feared to increase public unrest and were prone to being used unwisely by interested parties.
“The mathematical dynamic model used by some parties gives excessive predictions with very high errors and is recommended only to be used with caution for Indonesia,” said Fidelis.
Why does it have to be a PDDM model?
There are two main reasons why researchers chose the PDDM model in predicting the final potential for the Covid-19 pandemic in Indonesia.
1. Similarities to machine learning
According to Dedi, although the PDDM model is simple, it is able to provide excellent one-day prediction accuracy going forward.
He said it is comparable to the predictive ability of complex machine learning models such as artificial neural network models and other more sophisticated models.
2. Has more advantages
There are a number of advantages to this PDDM model, said Dedi, which are not shared by other models that were tested and developed previously.
Based on the PDDM model, the average error prediction error over the past 2 weeks is only 1.5 percent.
After testing forward predictions over the last 4 days since March 26, the resulting error has always been below 1 percent, a maximum of 0.9 percent and a minimum of 0.18 percent.
“Another advantage of the PDDM model is its ability to predict the worst times and the end of the Covid-19 pandemic in Indonesia,” Dedi said through his written statement
Covid-19 pandemic predictions from results of the PDDM model
With the PDDM model, Dedi said it was estimated that the maximum daily increase Covid-19 cases was around the second week of April, specifically around April 7-11, 2020. With an increase of approximately 185 patients per day, numbers are expected to continue to decline going forward.
Based on available data it is estimated that the pandemic will end approximately 100 days after March 2, around May 29, 2020. The total number of confirmed cases of Covid-19 patients is estimated at around 6174 cases.
“From mid-May 2020, the total increase in patients will be relatively small,” he said.
Accuracy of prediction results
For the accuracy of predicting confirmed Covid-19 cases through this model, the error percentage is often found below 1 percent.
However, researchers still recommend that returning home for Lebaran is not carried out and gatherings and activities during the month of Ramadan be cancelled. Government intervention through partial lockdown and strict social and physical distancing must continue until the pandemic effectively ends in early June 2020.
That is because, the accuracy of previous predictions are based on patient data up to March 26, 2020. In addition, it is assumed that there has been strict intervention from the government in place since the 3rd week of March, which has been successful.
Meanwhile, the effect of travelers from the big cities affected by Covid-19 since the new social – physical distancing regulations were enacted, has been assumed by researchers to be insignificant.
“This model also limits other external effects such as air temperature, population size, population density, etc., which are assumed to have no significant effect on the number of confirmed cases,” Dedi said.
This PDDM model will continue to be updated every day by the researchers so that the PDDM model predictions will truly reflect any changes from existing data.
Still, according to Dedi, study reports were based on an optimistic scenario. But this method can also be used to test various scenarios such as intervention and/or the influence of important external factors.
For example, this model can simulate the effects of an increase in Covid-19 cases in the last week of March (2020) due to the large number of travelers from large cities affected by Covid-19 to other regions.