Juan Ignacio Fulponi

Data partnership between Facebook, the University of Buenos Aires and the Inter-American Development Bank.

I started this project through a partnership with the University of Buenos Aires but we quickly realized that Facebook mobility data needed to be processed and analyzed in a more dedicated way. This is why he joined a consulting project at the IDB, working full time on the development of the dashboard and the reports for each of the cities involved. Machine learning and statistical mechanisms were used to estimate mean distances and origin and destination matrices from the raw data with R and the dashboard was created in a Shiny environment. This information is now part of CAF’s Urban Mobility Observatory. I have learned many skills through this and been I’m very proud of having this in my portfolio.

The Uber Team contacted me to make a empirical research to know if there is a statistical relationship between Uber trips and traffic jams in Buenos Aires.

During this project, I have used a lot of geocomputational tools in R and SQL to homogenize both Uber and Waze spatial databases. Once I had the final dataframe, I built a xgboost model that predicts in which area of the region the jam is. This demostrates the structural differences between the City and the suburbs. Then, it is compared with the behavior of Uber trips throughout the entire region with various statistical tests such as correlation, logistic regressions and other machine learning classification models. The study concludes that there is no statistical relationship between Uber trips and congestion in Buenos Aires.

This project was created in conjunction with the "Socioeconomic Vulnerability Assessment" research group of the School of Economics of the University of Buenos Aires and with the support of the Massachusetts Institute of Technology (MIT).

The model uses machine learning to predict trend production under a scenario of normal climatic variability, from which extreme deviations in yields or area losses are evaluated that are consistent with extreme climatic events, such as severe to extreme droughts and floods affecting large areas. From this point, the variables of production and harvest loss can be geolocated and an analysis dashboard of the final results can be created.

Another data science project.

In the middle of the pandemic I was involved in this lovely little project in which I joined a group of engineers to do market analysis and project valuation for the implementation of electric buses in Panama. Some tasks I have done:

1) Comprehensive economic analysis linked to transport and data issues.

2) Elaboration and processing of large databases.

3) Composition of dashboards for tracking electric bus routes in Panama in Shiny.

4) Machine Learning in R.

A lot of times we create some solutions for problems that not exist (yet) because we need to keep the brain in movement.

I love the governance of public data. I am very interested in how governments or big companies generate public information that can be processed, transformed and contextualized. In Argentina we have tons of public data but only a few “data communicators” or… data influencers? Social networks are not only for memes. I love to share some graphs or analysis of topics I like. For example, I have made a Twitter bot that tweets every hour the status of the bus system in Buenos Aires. The bot compiles public information from the Buenos Aires City Government and generates some interesting variables such as quantity of buses that are circulating above max speed. This data can be viewed in the Shiny dashbard too.