Fig 1.
The locations of Warsaw and Łódź in Poland, WGS-84 coordinate system.
The base map created using USGS National Map Viewer (https://www.usgs.gov/tools/national-map-viewer), shared under the CC BY 4.0 licence (https://creativecommons.org/licenses/by/4.0/deed.en).
Fig 2.
Visualization of BDOT10k for the Warsaw area (a), population density in Warsaw based on the GHS-POP model (b), distribution of bike stations in Warsaw (c) and Łódź (d) on an OpenStreetMap (OSM) base map, WGS-84 coordinate system.
The base map in panel a created with Quantum Geographic Information System (QGIS) software using Polish National Geoportal (https://www.geoportal.gov.pl/pl/dane/baza-danych-obiektow-topograficznych-bdot10k), shared under the CC BY 4.0 licence (https://creativecommons.org/licenses/by/4.0/deed.en) on basis of the permission granted by the owner of the data (Head Office of Geodesy and Cartography). The base map in panel b created with Quantum Geographic Information System (QGIS) software using GHSL data (https://human-settlement.emergency.copernicus.eu/index_op.php), shared under the CC BY 4.0 licence (https://creativecommons.org/licenses/by/4.0/deed.en). The base map in panel c and panel d created with Quantum Geographic Information System (QGIS) software using OpenStreetMap (https://www.openstreetmap.org/), shared under the Open Database Licence (http://www.opendatacommons.org/licenses/odbl).
Fig 3.
A diagram illustrating the workflow of the methodology.
Fig 4.
Overlay of the primary field grid, bike routes, and bike stations in the study area, along with the results of the intersection analysis.
Fig 5.
Diagram visualizing the decision tree used in the research.
Fig 6.
Dependency of neural network training accuracy on the number of hidden neurons for combinations of one (a) or two (b) hidden layers.
Fig 7.
The neural network structure with the best accuracy parameters among those tested in the research: 2 hidden layers and 1 activation function.
Fig 8.
Learning curves for neural networks for cases: 3 (a), 4 (b), and 5 (c).
Fig 9.
Visualization of prediction results for cases: (a) – 1, (b) – 2, (c) – 3, (d) – 4, (e) – 5.
All base maps created with Quantum Geographic Information System (QGIS) software using OpenStreetMap (https://www.openstreetmap.org/), shared under the Open Database Licence (http://www.opendatacommons.org/licenses/odbl).
Fig 10.
Prediction results for the bike station network locations in Warsaw using ANN with linear regression and the tanh activation function.
The base map created with Quantum Geographic Information System (QGIS) software using OpenStreetMap (https://www.openstreetmap.org/), shared under the Open Database Licence (http://www.opendatacommons.org/licenses/odbl).
Fig 11.
Locations identified in the research as suitable for bike stations in relation to their actual placement in Łódź.
All base maps created with Quantum Geographic Information System (QGIS) software using Polish National Geoportal (https://www.geoportal.gov.pl/pl/dane/baza-danych-obiektow-topograficznych-bdot10k), shared under the CC BY 4.0 licence (https://creativecommons.org/licenses/by/4.0/deed.en) on basis of the permission granted by the owner of the data (Head Office of Geodesy and Cartography).
Fig 12.
Bike station locations relative to their surroundings: a) actual, b) predicted.
Both cases show that stations are located at intersections and within the city centre. Both base maps created with Quantum Geographic Information System (QGIS) software using Polish National Geoportal (https://www.geoportal.gov.pl/pl/dane/baza-danych-obiektow-topograficznych-bdot10k), shared under the CC BY 4.0 licence (https://creativecommons.org/licenses/by/4.0/deed.en) on basis of the permission granted by the owner of the data (Head Office of Geodesy and Cartography).
Fig 13.
Prediction results for the bike station network locations in the area of Rzeszów.
The base map created with Quantum Geographic Information System (QGIS) software using OpenStreetMap (https://www.openstreetmap.org/), shared under the Open Database Licence (http://www.opendatacommons.org/licenses/odbl).