segment line-up as time-series

[tiffanychu90]

segment speeds

• use gtfs_digest/merge_data.py concatenated time-series data for segments
• exploratory notebook checking how many variations of segments there are between the same 2 stops (stop_pair, which should indicate the same physical segment entity)
  • handling the shape variation within a day + standardizing route names over time -> majority of segments track well over time, much better than anticipated
• time-series has over 17 months! these are rough findings:
  • 95% rows are getting it right, 5% has dupes (speed counts are more than the dates, which means there are some duplicated counts)
  • about 55% of segment geometries stay stable across 17 months...85% have 1 or 2 geometries over this period.
  • exploring where there are 3+ variations, a lot of these are still getting at the same segment, little visible difference, except start/end points have slightly different lat/lon floats
• TODO: set up parameterized notebook to explore how those speeds look. might be able to produce a very aggregated value, now that we know the building blocks are stable.
  • need a similar heuristic for route name harmonization / standardization applied here to deal with the little segment variation that still occurs -- probably just select the most recent one is simplest
• Research Request - Segment stability in speed time-series / quarterly and yearly averages #1202

[github-actions]

nbviewer URLs for impacted notebooks:

• rt_segment_speeds/39_segment_stability.ipynb

[github-actions]

nbviewer URLs for impacted notebooks:

• rt_segment_speeds/39_segment_stability.ipynb

[github-actions]

nbviewer URLs for impacted notebooks:

• rt_segment_speeds/39_segment_stability.ipynb