colocationship 0.0.1

A light package for build and analyse co-locationship

co-locationship

"Co-locationship" as an Alternative to Social Ties in Predictability of Human Mobility.

The code was tested on Python 3.6.

Install

The dependencies package are shown in requirements.txt, also, you can run

pip install -r requirements.txt

After installing all dependencies clone the repository and do (inside the top directory):

pip install . 

This will install a copy of the code as a package. If you want to install a package that links to the cloned code, run

pip install --editable .

This makes changes to the source files in the cloned directory immediately available to everything that imports the package.

Anywhere on the system you can then import the package:

import colocationship as cl

Dataset

All processed datasets (Weeplaces, BrightKite, Gowalla) we used in this repo can be found in Google Drive.

Usage

Build a Co-Locationship Network

df_checkin = pd.read_csv("../dataset/weeplace_checkins.csv")
co_locationship = cl.Co_Locationship(df_checkin,
                                     mins_records=150,
                                     missing='-')
# generate placeid sequence
co_locationship()

# calculate basic information of co-locationship network
co_locationship.calculate_info()

# network filter, including minimum both quality control and contribution control, and minimum 10 alters requirement and ranking.
co_locationship.network_control(quality=True, 
                                contribution=True, 
                                num_alters=10, 
                                by=['userid_x', 
                                    'meetup',
                                    'N_previous'], 
                                ascending=[True, 
                                           False, 
                                           True],
                                N_previous=150)

# calculate details of co-locationship network, including cumulative cross-entropy and cumulative cross-predictability
co_locationship.calculate_network(verbose=False, filesave=True)

The code in Line 6, Line 9, Line 24 may takes a long time, so it is better to run them separately, and save these interim results.

Social relationship network

The similar code for social relationship, just slight differences.

df_friend = pd.read_csv("../dataset/weeplace_friends.csv")
social_relationship = cl.Social_Relationship(df_checkin,
                                             df_friend,mins_records=150,
                                             missing='-')

# given the same time unit, we can just the placeid sequence generated by co-locationship, otherwise, also run `social_relationship()` to generate placeid sequence again.
social_relationship.placeidT = placeidT

social_relationship.calculate_info()

srn_filtered = social_relationship.network_control(
                                contribution=True, 
                                num_alters=10, 
                                by=['userid_x', 
                                    'meetup',
                                    'N_previous'], 
                                ascending=[True, 
                                           False, 
                                           True],
                                N_previous=150,
                                freq='H')
# calculate details of co-locationship network, including cumulative cross-entropy and cumulative cross-predictability
social_relationship.calculate_network(verbose=False, filesave=True)

Similarly, the code in Line 9, Line 23 may takes a long time, so it is better to run them separately, and save these interim results.

Comparison of co-locationship and social relationship

# Create a comparsion class to compare these two networks
compare_v2 = cl.ComparisonNetwork([co_locationship,
                                   social_relationship],
                                  ['co-locationship',
                                   'social relationship'])

# print the number of shared users in both networks
compare_v2()

# Histogram plot comparison
compare_v2.plot_CE(l=8,w=5)

# RCCP comparison based on ranking
compare_v2.plot_errorbar(target='RCCP', l=8,w=5)

# Local similarity comparison
compare_v2.plot_similarity(local=True, l=5,w=5)

# Global similarity comparison
compare_v2.plot_similarity(local=False, l=5,w=4)

# Monotonicity statistical test
compare_v2.stats_test_monotonicity(target='RCCP')

# Consistency statistical test
consistency_test = compare_v2.stats_test_consistency(target='RCCP')

Details

Details please refer to example/example_weeplaces.ipynb.

Contributing

PRs accepted.

License

MIT © Zexun Chen