search

fangzhou-xie / rethnicity

predict ethnicity from names
https://fangzhou-xie.github.io/rethnicity/index.html
8 stars 2 forks source link

Sharing results #7

Open max-mapper opened 11 months ago

max-mapper commented 11 months ago

I did a fun little project predicting 2 million LA County parcel registered owner names with this library. The main issues here seem to be with Filipino and Armenian names

Screenshot_20230810_155812

Here is predicted ethnicity of owner name vs reported occupant ethnicity in 2020 census (majority at block group level)

Screenshot_20230807_214143

fangzhou-xie commented 11 months ago

Thanks very much for sharing your interesting results! It's my pleasure to see that my work to be used in practice.

BTW, what are the colors used in the second plot? It seems that the model predicts more "yellow" race/ethnicity than it should be, as compared to "green" one. Alas, the prediction error is not trivial.

max-mapper commented 11 months ago

@fangzhou-xie In this case I don't think it's prediction error, the area here is El Monte California and the plot on the left is based on census (i.e. owners and renters) but the plot on the right where I predicted the ethnicity is based on property owner data (so just owners, not necessarily occupents). I think it is showing that the green (asian names) are the majority owners but yellow (hispanic/latino names) are the majority in the census

fangzhou-xie commented 11 months ago

I see. Thanks for your clarification! This seems to be an interesting result.