rafjaa/kdd-br-2018

Winning solution of the Kaggle KDD BR 2018 machine learning competition

KDD BR Competition 2018

Winning solution of the Kaggle KDD BR 2018 machine learning competition: https://www.kaggle.com/c/kddbr-2018

Code: https://github.com/rafjaa/KDD-BR-2018/blob/master/solution.ipynb

TSI Team

• Rafael Alencar
• João Paulo de Melo
• Guilherme Domith

Competition description

This competition is part of the 2nd KDD-BR (Brazilian Knowledge Discovery in Databases competition) event, one of the joint activities of the 2018 editions of BRACIS, ENIAC, KDMILE, CTDIAC.

In this competition, you will predict palm oil harvest productivity with data provided by AGROPALMA. The dataset contains information about palm trees, their harvest dates, atmospheric data during development of the plants and soil characteristics of the fields where the trees are located in.

To preserve anonymity in the actual harvest productivity values, the data was scaled to the range [0, 1] while still maintaining the same distribution found in the original dataset.

The dataset includes soil information from SoilGrids, and atmospheric data from the ERA-Interim reanalysis dataset.