In Sep 9th, 2020 I wrote a post where I discussed if a machine learning model could predict if farmers reach a living income, and why this is important.
Results where promising! The model was correct 96.5% of the time (accuracy). However, since only 15% of the farmers in the dataset reached a living income, a model can be highly accurate just by guessing no one will reach it. …
In Sep 9th, 2020 I wrote a first post where I discussed if a machine learning model could predict if farmers (in Ghana and Côte d’Ivoire) reach a living income, and why this is important. In Dec 16th, 2020, I wrote a follow-up post where I presented a machine learning model tackling the same problem using only 9 variables and web app using the model.
This post is the technical companion. Here I go into the technical details behind the two posts above and explain my path from beginning to end.
The Living Income Community of Practice has been making a tremendous job in developing a more challenging standard of living reference value: a minimum amount needed for a household to have a basic, but decent, standard of living.
This is higher than the poverty line and therefore harder to achieve. However, we all must agree that smallholder farmers (who produce a large amount of the food we consume) should not be living in (extreme) poverty.