• It’s difficult to detect the impact of various factors on the poverty of a particular household.
  • Most countries don’t have the resources to conduct comprehensive & accurate surveys which results in many factors & households being overlooked.
  • Most poverty predictions are done using financial data which is not accurate for a variety of reasons.
  • A cheaper method is needed that can use other kinds of data such as aerial surveys etc.

What it does

  • Visualizes factors affecting poverty in detail
  • Financial data of the population is NOT required to predict the poverty levels of every household.
  • Poverty prediction can be done from data taken from other sources such as aerial surveys, education surveys etc.
  • Reduces cost & manpower reqd. for conducting analysis.
  • Easy analysis of all the factors impacting poverty.

How we built it

We performed the data exploration, feature engineering & ML modelling all using Python on the extremely useful Jupyter Notebook

Challenges & Accomplisments

  • The data had a lot of factors & we had to wrap our head around its significance and impact on our target.
  • We were able to make an ML model with good results.
  • We learned to use real world data to solve challenges for social good.

What's next for Household Poverty Predictor

  • We plan to try other alogrithms to increase the F1 score
  • We plan to tune some hyperparameters later on

Built With

Share this project: