/WorldBankDSCompetition

Data Science competition at the World Bank

Primary LanguageJupyter Notebook

WorldBankDSCompetition

Poverty levels are often estimated at the country level by extrapolating the results of surveys taken on a subset of the population at the household or individual level.

The surveys are incredibly informative, but they are also incredibly long. A typical poverty survey has hundreds of questions, ranging from region-specific questions to questions about the last time a participant bought bread. In order to track progress towards its goal, The World Bank needs the most efficient survey possible.

The competition, entitled "Pover-T Tests" was about predicting poverty at the household level by building a classification model. The strongest poverty predictors could be used by statisticians at The World Bank to design new, shorter, equally informative surveys. With these improvements, The World Bank can more easily track progress towards their ambitious and inspiring goal.

This is a team effort: I worked with Elie Gerschel and Pierre Salvaire on this competition.

More info on: https://www.drivendata.org/competitions/50/worldbank-poverty-prediction/page/99/

Final rank: 77/1969