[Data] Investigate routes with ratio of actual trips to scheduled trips greater than 1

[dcjohnson24]

Investigate routes with ratio > 1

There are some routes that have a ratio of actual trips to scheduled trips greater than one, and it would be good to know why.

Access the data

Jupyter Notebook

To access the data, run the notebook compare_scheduled_and_rt.ipynb. Add a cell at the bottom with %store summary and run it. The %store magic command allows you to share variables between notebooks https://stackoverflow.com/questions/31621414/share-data-between-ipython-notebooks.

Next, run the static_gtfs_analysis.ipynb. Add a cell at the bottom with %store -r summary and run it to read the summary DataFrame from the compare_scheduled_and_rt.ipynb notebook. Merge the summary DataFrame with the final_gdf GeoDataFrame from the compare_scheduled_and_rt.ipynb using summary_gdf = summary.merge(final_gdf, how="right", on="route_id")

Python

Run the following in an interpreter from the project root:

import pandas as pd

import data_analysis.compare_scheduled_and_rt as csrt
import data_analysis.static_gtfs_analysis as sga

summary_df = csrt.main()

gdf = sga.main()

summary_gdf = summary_df.merge(gdf, how="right", on="route_id")

Find routes with ratio > 1

To filter the rows with ratio > 1, use

ratio_over_one = summary_gdf.loc[summary_gdf.ratio > 1]
ratio_over_one.head()

A few things to look for:

• Duration of trips
• How many trips cross the hour boundary
• Number of routes with ratio > 1 after reaggregting data based on a different frequency e.g. daily, see [Data] Adjust realtime/scheduled data comparison to cover multiple time frames or averages (for example, average over entire period vs. avg by day vs. avg by hour across entire period) #12

[porouspaper]

on trips crossing the hour boundary - are we suspecting that this code is double-counting trips if the trip crosses an hour boundary? despite vid being aggregated as a set?

[dcjohnson24]

I think that's the code. I guess maybe vid is unique only for a given hour, but it could appear in another hour for the same trip. It does seem strange though.