Input Features to Model

[jonfroehlich]

Here's a brainstorm and ongoing discussion of features to use as input to AccessScore model. @manaswis and I have been talking about this a lot so I'm sure I missed something.

Project Sidewalk Data

• [ ] Curb ramps
• [ ] Missing curb ramps
• [ ] Surface problems
• [ ] Sidewalk obstacles
• [ ] Lack of sidewalks

Other data

• [ ] Elevation data (e.g., grade/steepness is important as is frequency of elevation change)
• [ ] Maybe open.gov datasources like the dc curb ramp dataset
• [ ] Weather?

Proximity to POIs

• [ ] Transit
• [ ] Libraries, hospitals, grocery stores, etc.

Another thing to think about is the side of the road that the problem exists on. This would particularly impact routing algorithms but could also impact our visualizations, I think.

[richardwmcgovern]

Regarding side of the road, perhaps this can be inferred by cross-referencing a geometry table (do you have street shapefiles? something related?). Streets are represented as lines with from and to nodes. With these you can have a consistent understanding of topological "right" and "left" sides, by checking if the point observation is respectively "under" or "above" a line pointed in the positive x-direction, or vice-versa if the to node is down or to the left of the from node.

Can I access your data to see what can be fiddled with?