Research Request - Use gtfs_segments base nearest neighbors function in pipeline

[tiffanychu90]

Complete the below when receiving a research request, and continue to add to this issue as you receive additional details and produce deliverables. Be sure to also add the appropriate project-level label to this issue (eg gtfs-rt, DLA).

Research Question

Single sentence description:

• Related to Research Request - Can we leverage gtfs-segments in our pipeline? #988

Detailed description:

• After digging into gtfs_segments package, we will use their create_segments to get our stop segments. When we apply that create_segments function, it takes quite a bit longer in geopandas, but if we wrap it with dask.map_partitions, it will be much faster to use.
• But also, create_segments relies on a nearest_points base function that uses scipy.spatial CKDTree that we can use more broadly.
• We can extend the use case in at least 4 areas (anytime we want to ask, where are the nearest x points to this other point location):

1. Eric's segments are 1,000 m (sometimes between stops), and we can still find the cutoffs within a shape and ask where the nearest vp is. --> Where are the nearest vp to these arbitrary 1,000m cutoffs?
2. Stop segments: All trips along a given shape, if there are variations, one shape variation is selected to stand in for all the trips that travel along that shape. This way, our stop segments are stable. But, if there are 2 variations (shape-stop_sequence-stop_id combinations) for a given shape_array_key, then not every stop is reflected for that trip, only 1 variation. --> Where are the nearest vp to these more-easily-aggregated-segment cutoffs?
3. Trip stops / rt_stop_times table: We would want to reflect each trip-stop's location and the speed it took to get there. This might be a contributing factor / conceptual bug in why not all stops are monotonically increasing. --> Where are the nearest vp to these every stop for that trip, get stop arrival, and derive speed.
4. Road segments: As is, we have trouble parsing out enough vp to traverse a given segment, especially when the bus makes a turn onto a different road. --> Where are the nearest vp to these arbitrary road cutoffs?

Deliverables

• Add functions, and/or apply functions in scripts to do use gtfs_segments.nearest_points in all portions of the analytics pipeline