Dynamic Eigenvector Centralities

The original code for this project was developed by Neela Avudaiappan. This version of the code is a fork from Grace Glenn's implementation, who fixed bugs and restructured the code.

Paper / Purpose

This code serves as an implementation of the work on calculating keywords and their emerging importance outlined in:

Neela Avudaiappan, Alexander Herzog, Sneha Kadam, Yuheng Du, Jason Thatcher, and Ilya Safro, " Detecting and summarizing emergent events in microblogs and social media streams by dynamic centralities", in Proceedings of the 2017 IEEE International Conference on Big Data, 2017

Results

The results in the paper have been replicated on the Boston dataset using time intervals of 60 and 15 minutes, located in boston_examples.

Requirements

1. All code is run in Python 3.6 (Anaconda 4.3.0)
2. Data to be processed should be stored in ordered text files (i.e., file1.txt, file2.txt, ... fileN.txt for N intervals, or some other numbered format.)
3. Text files should contain one-document (i.e., one tweet) per line

Usage

• Ensure all requirements are satisfied. The program can be run as follows.

# after repo has been downloaded
cd dynamic_eigenvector_centralities
pip install requirements.txt
python dec_main.py --input_folder /home/username/time_series_data/ --P 6 --output_folder /home/username/dec_results/

Files

• dec_main.py Runs the full algorithm to compute DEC values described in the
• dec_graph.py contains code for the graph logic of the algorithm
• dec_text.py contains code for preprocessing and cleaning the data
• break_files.py a useful script for dividing time-series CSV data