Awesome

GTFS functions

This package allows you to create various layers directly from the GTFS and visualize the results in the most straightforward way possible.

Update November 2023:

• Possibility to check the service_id for a given date:

parsed_calendar = Feed(gtfs_path).parse_calendar()

or if you want it already grouped by date:

date_service = Feed(gtfs_path).get_dates_service_id()

Update August 2023:

• Possibility to parse the GTFS for a specific date range.

feed = Feed(gtfs_path, start_date='2023-03-31', end_date='2023-04-04')

Update March 2023:

• Removed dependency with partridge. As much as we love this package and think it is absolutely great, removing a dependency gives us more control and keeps this package from failing whenever something changes in partridge.
• We treat the GTFS as a class, where each file is a property. See examples below to find out how to work with it. We hope this simplifies your code.
• Fixed and enhanced segment cutting. Shout out to Mattijs De Paepe
• Support to identify route patterns!! Check it out using feed.routes_patterns. Shout out to Tobias Bartsch
• The rest should stay the same.

Warning!

Make sure stop_times.txt has no Null values in the columns arrival_time and departure_time. If this is not the case, some functions on this package might fail.

Table of contents

• Installation
• GTFS parsing
• Stop frequencies
• Line frequencies
• Cut in Bus segments
• Speeds
• Segment frequencies
• Mapping the results
• Other plots

Python version

The package requires python>=3.8. You can create a new environment with this version using conda:

conda create -n new-env python=3.8

Installation <a class="anchor" id="installation"></a>

You can install the package running the following in your console:

pip install gtfs_functions

Import the package in your script/notebook

from gtfs_functions import Feed

GTFS Import <a class="anchor" id="gtfs_parsing"></a>

Now you can interact with your GTFS with the class Feed. Take a look at the class with ?Feed to check what arguments you can specify.

gtfs_path = 'data/sfmta.zip'

# It also works with URL's
gtfs_path = 'https://transitfeeds.com/p/sfmta/60/latest/download'

feed = Feed(gtfs_path, time_windows=[0, 6, 10, 12, 16, 19, 24])

routes = feed.routes
routes.head(2)

stops = feed.stops
stops.head(2)

stop_times = feed.stop_times
stop_times.head(2)

trips = feed.trips
trips.head(2)

shapes = feed.shapes
shapes.head(2)

Stop frequencies <a class="anchor" id="stop_freq"></a>

Returns a geodataframe with the frequency for each combination of stop, time of day and direction. Each row with a Point geometry. The user can optionally specify cutoffs as a list in case the default is not good. These cutoffs should be specified at the moment of reading the Feed class. These cutoffs are the times of days to use as aggregation.

time_windows = [0, 6, 9, 15.5, 19, 22, 24]

feed = Feed(gtfs_path, time_windows=time_windows)
stop_freq = feed.stops_freq
stop_freq.head(2)

Line frequencies <a class="anchor" id="line_freq"></a>

Returns a geodataframe with the frequency for each combination of line, time of day and direction. Each row with a LineString geometry. The user can optionally specify cutoffs as a list in case the default is not good. These cutoffs should be specified at the moment of reading the Feed class. These cutoffs are the times of days to use as aggregation.

line_freq = feed.lines_freq
line_freq.head()

Bus segments <a class="anchor" id="segments"></a>

Returns a geodataframe where each segment is a row and has a LineString geometry.

segments_gdf = feed.segments
segments_gdf.head(2)

Scheduled Speeds <a class="anchor" id="speeds"></a>

Returns a geodataframe with the speed_kmh for each combination of route, segment, time of day and direction. Each row with a LineString geometry. The user can optionally specify cutoffs as explained in previous sections.

# Cutoffs to make get hourly values
speeds = feed.avg_speeds
speeds.head(1)

Segment frequencies <a class="anchor" id="segments_freq"></a>

segments_freq = feed.segments_freq
segments_freq.head(2)

Map your work <a class="anchor" id="map_gdf"></a>

Stop frequencies

# Stops
from gtfs_functions.gtfs_plots import map_gdf

condition_dir = stop_freq.dir_id == 'Inbound'
condition_window = stop_freq.window == '6:00-9:00'

gdf = stop_freq.loc[(condition_dir & condition_window),:].reset_index()

map_gdf(
  gdf = gdf, 
  variable = 'ntrips', 
  colors = ["#d13870", "#e895b3" ,'#55d992', '#3ab071', '#0e8955','#066a40'], 
  tooltip_var = ['min_per_trip'] , 
  tooltip_labels = ['Frequency: '], 
  breaks = [10, 20, 30, 40, 120, 200]
)

Line frequencies

# Line frequencies
from gtfs_functions.gtfs_plots import map_gdf

condition_dir = line_freq.direction_id == 'Inbound'
condition_window = line_freq.window == '6:00-9:00'

gdf = line_freq.loc[(condition_dir & condition_window),:].reset_index()

map_gdf(
  gdf = gdf, 
  variable = 'ntrips', 
  colors = ["#d13870", "#e895b3" ,'#55d992', '#3ab071', '#0e8955','#066a40'], 
  tooltip_var = ['route_name'] , 
  tooltip_labels = ['Route: '], 
  breaks = [5, 10, 20, 50]
)

Speeds

If you are looking to visualize data at the segment level for all lines I recommend you go with something more powerful like kepler.gl (AKA my favorite data viz library). For example, to check the scheduled speeds per segment:

# Speeds
import keplergl as kp
m = kp.KeplerGl(data=dict(data=speeds, name='Speed Lines'), height=400)
m

Segment frequencies

# Segment frequencies
import keplergl as kp
m = kp.KeplerGl(data=dict(data=seg_freq, name='Segment frequency'), height=400)
m

Other plots <a class="anchor" id="plotly"></a>

Histogram

# Histogram
import plotly.express as px
px.histogram(
    stop_freq.loc[stop_freq.min_per_trip<50], 
    x='frequency', 
    title='Stop frequencies',
    template='simple_white', 
    nbins =20)

Heatmap

# Heatmap
import plotly.graph_objects as go
dir_0 = speeds.loc[(speeds.dir_id=='Inbound')&(speeds.route_name=='1 CALIFORNIA')].sort_values(by='stop_sequence') 
dir_0['hour'] = dir_0.window.apply(lambda x: int(x.split(':')[0]))
dir_0.sort_values(by='hour', ascending=True, inplace=True)

fig = go.Figure(data=go.Heatmap(
                   z=dir_0.speed_kmh,
                   y=dir_0.start_stop_name,
                   x=dir_0.window,
                   hoverongaps = False,
                   colorscale=px.colors.colorbrewer.RdYlBu, 
                   reversescale=False
))

fig.update_yaxes(title_text='Stop', autorange='reversed')
fig.update_xaxes(title_text='Hour of day', side='top')
fig.update_layout(showlegend=False, height=600, width=1000,
                 title='Speed heatmap per direction and hour of the day')

fig.show()

Line chart

by_hour = speeds.pivot_table('speed_kmh', index = ['window'], aggfunc = ['mean','std'] ).reset_index()
by_hour.columns = ['_'.join(col).strip() for col in by_hour.columns.values]
by_hour['hour'] = by_hour.window_.apply(lambda x: int(x.split(':')[0]))
by_hour.sort_values(by='hour', ascending=True, inplace=True)

# Scatter
fig = px.line(by_hour, 
           x='window_', 
           y='mean_speed_kmh', 
           template='simple_white', 
           #error_y = 'std_speed_kmh'
                )

fig.update_yaxes(rangemode='tozero')

fig.show()