Detecting country of residence from social media data – Python scripts and mock data

This repository provides supplementary information for the article Detecting country of residence from social media data: a comparison of methods.

The methodology has been further developed in the BORDERSPACE project at the University of Helsinki. Updated code will be made available via https://github.com/DigitalGeographyLab/

Citation

To cite any material in this repository and the related journal article, please use the following citation:

Heikinheimo, V., Järv, O., Tenkanen, H., Hiippala, T., Toivonen, T. (2022). Detecting country of residence from social media data: a comparison of methods. International Journal of Geographical Information Science 36(10): 1931-1952. DOI: 10.1080/13658816.2022.2044484

Source data

• Global posting history covering years 2010-2016 of social media users who visited Kruger National Park in 2014
  • Based on publicly available data collected from the Instagram API in spring 2016
  • Each point has information of the related administrative areas (Contintents, subregions & countries)
  • This repository contains de-identified mock demo data in demo_data folder. Geotags, timestamps and user identifiers are fake in this layer, but it allows testing the origin detection techniques.
• Global regions -layer
  • Based on GADM 2019 (Database of Global Administrative Areas).
  • Modified by the authors on the sub-regional level.
  • This layer is not included in this repository due to GADM license.
  • We welcome interested scholars to contact the authors of the related paper for more details.

Results in the manuscript are based on data from 1375 users (33 % sample included in the expert assessment)

Origin detection approaches

• BASIC (no hierarchy)
• HIERARCHICAL ( Continent --> Subregion --> Country)

APPROACH	Measuring technique	Definition of home location	Associated Tool / Script
SIMPLE	number of posts	Country with most posts	BASIC HIERARCHICAL
SPATIAL	mean center	Country where average centre point is located	Mean center in: BASIC HIERARCHICAL
SPATIAL	median center	Country where median center is located	Median center in: BASIC HIERARCHICAL
	center of SD ellipse	Country where centre point of Standard Deviational Ellipse is located	Ellipse centroid in: BASIC HIERARCHICAL
	center of SD circle	Country where centre point of Standard Distance Circle is located	Circle centroid in: BASIC HIERARCHICAL
	clustering	Country where centre point of most significant cluster is located	DBSCAN: BASIC HIERARCHICAL
TEMPORAL	max visit length	Country which has longest stay time between last – first post date	MaxTimedelta in: BASIC HIERARCHICAL
	frequency by months	Country with max frequency by active months. If two or more countries have equal frequency, then country with most posts is chosen	MaxMonths in: BASIC HIERARCHICAL
	frequency by weeks	Country with max frequency by active weeks. If two or more countries have equal frequency, then country with most posts is chosen	MaxWeeks in: BASIC HIERARCHICAL
	frequency by days	Country with max frequency by active days. If two or more countries have equal frequency, then country with most posts is chosen	MaxDays in: BASIC HIERARCHICAL

Results

Result files used in the final analysis in valid_results -folder:

• results_combined_by_user.csv
• origin_assesment_expert_1_expert_2.csv
• f1-scores-micro_macro.csv
• spearman_stats.csv

Results based on the demo data can be found in folder demo_results

Jupyter notebooks for plotting the result tables and figures:

Note, all input data for running these notebooks is not readily available in this demo repository.

• F1-scores
• expert agreement
• Spearmann correlation
• slopegraph
• Maps