Code on GitHub! More from CoronaCidades

The simulator was built to assist the decision making of public managers in the fight against the COVID-19 pandemic, foreseeing the overload of the capacity of hospital resources. We adapted the SEIR model from Alison Hill (2020) and used the data from Brazilian states’ daily reports collected by Brasil.IO volunteers (including me!), and DataSUS municipal health resources.

This project is one of many tools of the CoronaCidades initiative, by Impulso-gov and partners, to guide municipality’s policies with the advancement of Covid-19 - and it couldn’t be done without the help of volunteers, including João Carabetta, Ana Paula Pellegrino, Diego Oliveira, Gabriel Saruhashi and Luiz Felipe Costa. The project is still ongoing - and hopefully, it will serve good in this time of need.

Code on GitHub! OpenData Day (pt)

The project aims to build a freshman dropout predictor for high schools in Rio de Janeiro state using public data. This work was presented on Cerveja com Dados: Open Data Day.

According to a study by Insper, one in every 4 young people aged 15 to 17 interrupt their studies at this stage. Several dropout recovery programs have emerged in recent years in different states and municipalities. The current public policy acts after the student leaves the educational system. My goal was to measure the likelihood of a student to dropout, creating a vulnerability indicator that could be used in a preemptive public policy.

The data sourced from the Brazilian School Census (2016/2017), crossed with schools and favelas geolocation. To contemplate the grouping structure of students within schools, multilevel regression and mixed random forest models were tested. Some of the findings were that student’s age played a major role in prediction - as a baseline, it had an AUC of .6 -, and also other features such as the use of public transportation had a positive effect of reducing dropout.