First pass at feature selection (Aug 1 meeting) - Githubissues

dime-worldbank / big-data-poverty-estimation

Develop machine learning models for poverty estimation

25 stars 7 forks source link

First pass at feature selection (Aug 1 meeting) #3

Closed jtanwk closed 4 years ago

jtanwk commented 5 years ago

Data prep:

[x] Merge on income data with satellite data
[x] Create indices from all possible pairs of Landsat 7 bands

Analysis:

[x] Run data through small-grid ML pipeline
[x] Try daytime-only, nighttime-only, and combined feature sets
[x] Check which features are most predictive of household income