bikesharing-factors-evaluation

Seminar paper analyzing factors driving bikesharing demand in Hamburg.

I created this analysis in a Computational Social Science Seminar. It uses a dataset of bikesharing trips with start and stop station in Hamburg from "Call A Bike", a station based bikesharing system from Deutsche Bahn AG. The dataset is available here.

The README gives a short overview over the project, the detailed description can be found in reporting/2020_06_seminar_paper_css_classen.

Requirements

The project was conducted in R using the following packages, requiring an R Version > 3.4:

data.table
tidyverse
lubridate
sp
sf
ggplot2
leaflet
glmnet
Hmisc
lattice
MASS
pscl
lmtest

Analysis

The project uses a negative binomial regression with zero-inflation to account for the large number of zero-trips in the bikesharing network (> 90 % of all station combinations had zero trips). Model estimation can be found in 04_analysis.R.

The target variable is number of trips between station-pairs per day.

The input features are created in the 02_... RScripts and contain the following:

• Number of Points of Interest (POI) of different categories within a radius of 400 m of origin (O) and destination (D) station, categories are leisure, transport, eudcaiton, health
• Distance to closest POI for O and D
• Number of other stations within a radius of 400 m of O and D, shortest distance to closest station
• Distance to city center for O and D
• Percentage of landuse type around 400 m of O and D
• Percentage cycle way on shortest path
• Trip length in minutes (based on estimated shortest path)

03_combine_features.R merges all features into one dataset.

The results showed decrease in bikesharing trips with increased trip length and smaller population, as well as a positive impact of percentage of cycle ways on demand.