With the use of optimization and control models and neural networks, in conjunction with epidemiological and economic models, researchers will determine the ideal isolation rate by taking into account the social, economic, and demographic characteristics of municipalities

The adoption of practices involving quarantine and social distancing, which looked promising for containing the Covid-19 pandemic in European and Asian countries, was found to be difficult to implement in Brazil, due to extensive territorial distances and enormous social and economic inequalities. In practice, this difficulty comes to the forefront when the population “interrupts” social distancing, thus opening a gap for new waves of epidemics and a greater unfolding of economic impacts.

In an attempt to reverse this situation, Brazilian scientists, from four public universities, came together to solve the problem by positing models of optimization and control, as well as neural networks, in conjunction with epidemiological and economic models.

“Our objective is to find the optimal isolation rate, which is appropriate for each city, in order to put the brakes on the advance of the pandemic and lessen the adverse effects on the economy,” explained project Coordinator, Emílio Carlos Nelli Silva, Professor of the Department of Mechatronics Engineering and Mechanical Systems of the Polytechnic School of the University of São Paulo (Poli-USP) and member of the FAPESP Shell Research Centre for Gas Innovation (RCGI). “The project is a case study of the municipalities of the State of São Paulo, but it is expected that the results of the study can be replicated anywhere else in Brazil,” he adds.

Such a challenge requires the work of researchers with different types of expertise. Involved from Poli-USP and the RCGI, are Professors Celma de Oliveira Ribeiro (economic models and neural networks), Cláudio Oller do Nascimento (propagation), Emílio Carlos Nelli Silva (regional contingency optimization), Julio Meneghini (simulation), Oswaldo Luiz do Valle Costa (optimal control), and researcher Sara Malvar (computational learning). From the State University of Rio de Janeiro (UERJ), Americo Barbosa da Cunha Junior (dynamics of pandemics); from the Federal University of Rio de Janeiro (UFRJ), Roberto Ivo de Lima Filho (economic models); and from the Federal University of São Paulo (UNIFESP), André Marcorin de Oliveira (control).

Optimal isolation – One of the crucial points of the study is to understand how the pandemic affects the work force and vice-versa. To that end, models will be used that trace the evolution of the pandemic, such as SIR, in conjunction with economic models, by using data from a variety of open sources, like the IBGE and Government Ministries. “Our focus is on obtaining a scalable decision model that can be applied geographically at the micro and macro levels, simulating scenarios that indicate the optimal isolation rate,” explains the Project’s Deputy Coordinator, Celma de Oliveira Ribeiro, Professor in the Production Engineering Department of Poli-USP.

Mobility factor – Associated with this model, another will be created to forecast how the pandemic spreads within a city or between municipalities. “In order to design the way in which the coronavirus is propagated, we will use a model with a set of chemical reactors to predict the flow of contaminated people. In this case, we would apply mobility data and the number of cases of the disease,” concludes Claudio Oller do Nascimento, Professor in the Chemical Engineering Department of Poli-USP.

Taking the Metro as an example, it would be possible to estimate the number of infected persons at the beginning of the line and how and to what extent the disease spreads at the destination area. “The same calculation can be applied to municipalities, using a variety of data, including highway traffic,” he explained.

Common characteristics – Another aspect of the project is dedicated to gathering characteristics that would be common in the municipalities affected by the pandemic. “We are gathering social, economic, and demographic data about all of the municipalities in the State of São Paulo, so as to correct these characteristics in line with the number of cases of the disease and the resulting deaths,” explains researcher Sara Malvar, who is a specialist in machine learning.

“By using an algorithm called gradient boosting (a machine learning method), it is possible to know what the most relevant variables are – birth rate, employment rate, number of hospital beds, energy consumption, basic sanitation, number of vehicles, GDP, per capita income, etc. – for predicting new cases of the disease, as well as deaths,” Malvar states.

“Once those variables are identified, it is possible to make a heat map, showing more cases. Therefore, it would be possible to segment the municipalities into clusters, indicating the same appropriate distancing rules for a group of cities with similar characteristics,” she adds.

Uniting the models – When the Project is finished, the researchers want to have answers for the complex issues of this pandemic: in the short and medium term, what would be the most promising strategies for reducing the impact of Covid-19 on health and on the economy? Once the ideal isolation strategy has been put together, what would be its effect on the economy? What investments should be made and with what volume? In what way can those investments ensure economic growth, in the case of new waves of the pandemic?

“And how does the evolution of the pandemic interfere with the work force and, as a result, with the growth of the nation’s production?” adds Emílio Carlos Nelli Silva. “These answers are of the utmost importance to address the crisis of this pandemic. The answers are needed, so that public officials can have resources at hand for elaborating effective public policies,” he concluded.