nodes_all_trips.csv and links_after_rejection.scv are the ODs used as potential demand. These are a filtered version afthe the rejection sampling method described in the paper. The original data was the result from a previous work (​Florez et al., 2017​)

step_1_parse_trajectories.ipynb parses raw GPS data of the BIKO app.

step_2_filter_mapmatch.ipynb obtains the map matching for the above GPS data. It parses the trajectories, analyzes the travel speed and duration as well as the departures of each trip. We define our potential demand as all the trips that reproduce the same travel distance than the observed in the BIKO app.

step3_potential_demand.ipynb calculates the shortest path between nodes from our Potential Demand. As inputs, it uses nodes_all_trips and links_after_rejection csv files. It also generates the flow network in the OSM network which corresponds to the file pair_potential_bike20190401.csv

step4_percolation.ipynb perform the percolation on a network made up by (1) potential flow network + (2) the existed bikepaths in Bogota.

step_5_identify_bike_faciliites.ipynb after getting the final network it determines which edges do not have an existing or projected bicycle facility on them

The latter file communities_table.ipynb is the notebook to generate the figure 7 from the paper.