Our team has shown consistent results when competing against thousands of people in different competitions, since 2016.
Financial Forecasting Challenge 2018 – G-Research
In this competition, hosted by the company G-Research, our team got the 2nd position. The challenge here was to predict the return of a financial data series, based on a database provided by the company.
Knowledge Discovery in Databases 2018 (KDD)
Our team reached the 2nd position in this competition, in 2018. Using a database provided by Agropalma, the challenge was the prediction of oil palm harvest productivity based on information related to palm trees, their harvest dates, atmospheric data during their development and soil characteristics of the fields where the trees were located in.
Data Science Game 2018
In this worldwide competition, our team reached the 4th position. During the finals, the teams faced challenges mixing sequential data and unstructured information to help and e-commerce industry to transform its business.
Knowledge Discovery in Databases 2017 (KDD)
In 2017, our team got the 1st position in this competition, which happened within the 6th Brazilian Conference on Intelligent Systems. The goal here was the classification of images of meteors (popularly known as shooting stars) captured by a monitoring station installed at the University of Vale do Paraíba astronomy observatory. The database was composed by images of meteors, but also by wrong captured images, such as birds, insects, planes and even lightning. Another difficulty was the classification of images in which the weather was cloudy, rainy or with some varying stars.
Kobe Bryant Shot Selection 2016
In this competition, launched in 2016 to celebrate Kobe Bryant retirement from NBA, our team got the 1st position (check “public leaderboard”). Using 20 year of data on Kobe’s swishes and misses, the competitors had to predict which shots would find the bottom of the net.
The Winton Stock Market Challenge 2016
Our team reached the 2nd position in this competition, hosted by Winton. The challenge here was the prediction, despite all the noise, of stock returns (intra and end of day), given a historical stock performance and masked features.
Improve and Go 