---

d4dttimes

This repository contains the python code for extracting travel times between different locations based on anonymized call detail record data.

The format of the input data is explained in http://arxiv.org/abs/1407.4885 (tower level mobility data). The original research data was provided by Orange and Sonatel in conjunction with the Data for Development Challenge Senegal (D4D 2014, http://www.d4d.orange.com/en/home/).

This repository contains the 'minimal set' of tools for extracting travel times from CDR data provided in the above format. (This code is a cleaned, stripped down version of the actual research code.) For more advanced pipelining of data, you may contact the author of this repository.

Author:

Rainer Kujala, Rainer.Kujala [at sign) aalto.fi

Contents:

aux.py

A couple of helper functions.

cell_tower_groups_from_coords.py

Function how to group cell towers given a location and a radius.

comp_tower_tower_travel_times.py

Functions for computing all travel times between two locations.

compute_on_road_distances.py

Functions for computing the on-road distances between locations.

dataio.py

Helper functions for loading data.

ttime_estimator.py

Estimates travel times given the travel times obtained using `comp_tower_tower_travel_times.py`

A function for computing bootstrap estimates.

example_pipeline.py

Simple script to show how these should be used in combination.

Testing

Test modules:

test_aux.py
test_comps.py

Dependencies

This package depends on the following packages.

• The SciPy stack: (Numpy, Scipy, Matplotlib)
• geopy (used for computing distances between coordinates)

City coordinate data (cities_coords_senegal.csv) has been obtained from http://www.tageo.com/index-e-sg-cities-SN.htm