Chwiggy / BA-Thesis-App

Multi-Temporal Reach

This project is an attempt at creating a tool to detect gaps in public transit coverage from both a network analysis perspective as well as an intersectional perspective. It is part of a larger thesis project with a not-yet-started write-up here

Installation

Clone this repository. Move into the folder And run

docker build -t thesis .

Usage

Setup

Create a folder with the following contents

folder	contents
gtfs	gtfs data either as subdirectory or as zip archive
population	population data from the Global Human Settlement Layer
output	output will be saved here as a GeoJSON

Execution

Run the following docker command with

docker run -v $DATAFOLDER$:/data thesis -g $GTFSPATH$ place

• DATAFOLDER needs to be an absolute path to a data folder.
• GTFSPATH needs to be /data/ + the relative path to the desired gtfs file within your DATAFOLDER.
• place is a location string that will be passed to the nominatim API.

For example:

docker run -v /home/emily/thesis_BA/data/:/data -g /data/gtfs/2023_rnv_gtfs.zip thesis Wiesloch

docker run -v $HOME/data:/data --rm --security-opt=seccomp=unconfined --workdir=/ thesis2 -g /data/gtfs/2024-02-19_Germany.zip Heidelberg

Things to keep in mind

Esoteric WSL 2 and VPN shit

Get-NetAdapter | Where-Object {$_.InterfaceDescription -Match "Cisco AnyConnect"} | Set-NetIPInterface -InterfaceMetric 4000

Get-NetIPInterface -InterfaceAlias "vEthernet (WSL)" | Set-NetIPInterface -InterfaceMetric 1

Requirements

Needs a recent Docker installation

Questions

Can closeness centrality map on to car and public transport differences Can closeness centrality work as an indicator for temporal public transport variabilty within cities? Why?

TODOs

• population data and lorenz curves
• temporal analysis: many departure times a day, compare results: use case ebay classifieds and friends
• think about equalizing of closeness centrality
• final run on server
  • optimise script for server usage
  • difference between 10th and 90th percentile as a measure of "turning up and wait"-abilty vs a need to plan
  • run analysis for heidelberg + surrounding area on server

New repository

• create new repo for analysis
• how much planning over the area
• compare different profiles: use classification?
• pick different representative cells with detailed analysis
  • from one cell to any cell over time
  • detailed routing for outlier connections
• group by neighbourhoods

Optionals

• compare different cities
• compare to car but with added door to door delays for cars
• compare methods for schools closeness centrality and isochrone
• open questions!
• compare cities with different layouts