In the last months, our team has followed the trends of multiple waves in several of the countries analyzed, as well as the presence of weekly seasonality in the states of Brazil. These characteristics led to the implementation of different extensions of our model for each state and country on our list, which are actually the topics of the last posts here on this blog.

However, with the natural evolution of the complexity of the implemented models, many of these models have required more time to finish the analyzes. Recall that we currently provide daily predictions for a list of 46 countries, in addition to Brazil and its 27 states, with each region having two series (confirmed cases and deaths) adjusted separately. Even with all our efforts to optimize and parallelize the codes, it has not been uncommon to find several of these models needing more than 24 hours to finish.

With that, we inform that as of today, our forecasts will be updated only twice a week, on Mondays and Thursdays.

With the evolution of the Covid pandemic, several characteristics of its development have been observed throughout the world and in the Brazilian states. Since its creation, the CovidLP project has set out to offer appropriate prediction models, based entirely on observed data, for Brazil and the world. However, given the various developments of the pandemic, the project's capabilities - both in terms of personnel and in terms of computation - impose some decisions that had to be taken to ensure that the project develops in an appropriate manner, following as far as possible the quality standards determined by the group.

The project monitoring group observed a pattern of weekly seasonality and the arrival of the second and third waves of the pandemic in several countries around the world. These characteristics had not been observed previously. Adequacy to these patterns generates a demand for the project far beyond that achievable by the team, due to the profusion of possibilities arising from the numerous combinations of the model's characteristics (number of waves, presence of seasonality in each wave, type of seasonality of each wave, etc.) . This makes it impossible to identify and implement at the moment the most appropriate combination for each country at the global level.

Due to this essentially operational difficulty, it was decided that the main focus of the most complete modeling will be given to Brazil and its states. This will already give users a good idea about the possibilities that the project provides. In the other countries of the world, the multiple wave model will be implemented without including seasonality, which will provide a more comprehensive but admittedly less specific overview. Fortunately, the project has its software (Package 'PandemicLP' available for installation) and code open to all, allowing the most interested user to make use of all the project's features in addition to those available in the app.

Finally, vaccination programs have recently started around the world. We see this fact with great optimism and, we know that in case of success, it will drastically affect the behavior of the evolution of the pandemic around the world, requiring the incorporation of new characteristics in our models to adapt to these future events. Thus, we remain attentive to all these changes.

Since October 31st, 2020, it has been implemented a prediction model for cases and deaths which allows the adjustment of three waves for some countries of our application. The adoption of this model is being made as this trend is being detected in the observed data series.

As of today, the predictions are available considering a model adapted to the possibility of two waves for some countries (including the countries we mentioned here).

Our team continues to monitor the results of this new model, and also to evaluate other possible adjustments in the methodology that may become necessary.