search

TomorrowsCities / tomorrowscities

MIT License
3 stars 1 forks source link

Loose geospatial connection between individual and schools #35

Closed hkayabilisim closed 1 month ago

hkayabilisim commented 5 months ago

The left building is a school while the right one is a residential building. Even if the building can access to school via road network, metric2 (number of children lost access to school) is 5. This is very strange because the the building is next to school.

After investigating the issue, it turns that there five five children in the building but none of them goes to this school. The children are associated with other schools in the area. And because of the flooded roads, there is not road access, so metric2 becomes nonzero.

This is the main reason of too large metric2. What can we do about this @PrashantR-79 ? The data is provided in https://github.com/TomorrowsCities/tomorrowscities/issues/33 by @jamaldabbeek .

image
PrashantR-79 commented 4 months ago

This situation arises because currently students are assigned to schools randomly. If students were assigned schools based on nearest neighbourhood, this situation might not arise. But this issue is a bit tricky to solve - do we assume that all students go to the nearest school? or do we assume that students may have preference for other schools farther from their homes?

hkayabilisim commented 4 months ago

Thanks for the explanation. I agree that achieving a "realistic" student-school association is very tricky and can't be done easily. But at least we figured out exactly why the metric2 can be very large even if road network is slightly damaged.

hkayabilisim commented 1 month ago

Re-implementing student-school association in data generation is left to the data generation team.