Development and contribution

This project is no longer under active development. Please refer to https://www.floodtags.com for latest developments on the floodtags product.

FloodTags Social Network Analysis

This repository contains the scripts required for performing a basic Social Network Analysis and influencer analysis on FloodTags data. The entire analysis can be run from a single command line command, but the individual scripts can also be used independently and tied into existing workflows. The files contained do the following:

• SNA_execute: The main launch file. Launch from terminal with python SNA_execute.py <starting_date> <ending_date> . Format for the dates is: YYYY-MM-DD. This will run all three major scripts in turn through a single command line.

• api_request: Contains the logic to request data from the Floodtags API and restructure it for further processing. Outputs a pandas dataframe as csv.

• mentions_to_edgelist: Contains the logic to convert dataset from raw tweets to a directed graph edge list based on @mentions, required for SNA. Outputs edge list as csv.

• SNA_script: Contains the logic of the Social Network Analysis and feature calculations. Outputs 4 csv files with the top 50 users in each category.

• csv files: Example output files of the SNA script based on real FloodTags data.

Installation:

1. Create Virtual Environment on server
2. Install dependencies with pip install -r requirements.txt
3. Run file in directory using python SNA_execute.py <starting_date> <ending_date> <database>. The arguments are mandatory, no defaults are provided. Format for the dates is: YYYY-MM-DD.
4. All .csv files will be created in the directory containing the top 50 of each of four user statistics

Interpretation of results

in-degree: the number of times a user is mentioned by other users

out-degree: the number of times a user has mentioned other users

page-rank: the degree by which a user was mentioned by other users, placing more weight on mentions from other high ranked users

betweenness centrality: the relative importance of a user in the network in terms of reach. Without these users there is no communication.

Further considerations

• Results are better over larger data sets.

• It is recommended to perform this analysis at the following times: once a month, once a week during rainy season, and after every major flood event.