Social and Behavioral Networks

Italian Referendum 2016

Authors

• Pedro Magalhāes Bernardo
• Amir Abbasinejad

How to Run

1. Clone the repository

Use the download link above or run the command:

git clone https://github.com/pedromb/sbn-italianref.git

2. Provide the datasets

Under /src/main/resources include the following files:

• Official_SBN-ITA-2016-Net.gz - File with connected graph.
• stream/ - Folder with the tweet data. The files are assumed to be compressed (gzip).

3. Run

You can execute using maven, just go to the root folder of the project and run:

mvn compile
mvn exec:java

P.S.: The first execution might take awhile since it has to generate the Lucene index. The indexing process can take between 30 min and 1h (or more depending or your machine) to be finished. Running the program for a second time should be faster. Also, the result of running the KPP-NEG algorithm is included here so it doesnt have to be ran again, since this process can take more than 1 day.

4. Graphs on report

The graphs on the report were generated using a Python script. You can find the IPython notebook for that on generate_graphs.ipynb

5. Files Generated

The following files are generated after executing:

1. lucene_index/ - The folder with the Lucene index to manage the data.

2. temporal_analysis/ - The folder with data generated for the temporal analysis part.

   2.1 temporal_analysis/tweets_distribution.csv - Distribution of tweets by support (YES/NO) over time.

   2.2 temporal_analysis/clusters.csv - Components of each cluster for YES and NO supporters.

   2.3 temporal_analysis/graphs/no - Graph, largest connected components and k-core for each cluster of NO supporters

   2.4 temporal_analysis/graphs/yes - Graph, largest connected components and k-core for each cluster of YES supporters

   2.5 temporal_analysis/k_core_timeseries_no.csv - Distribution of tweets over time for the k-core components of each cluster of NO supporters.

   2.6 temporal_analysis/k_core_timeseries_yes.csv - Distribution of tweets over time for the k-core components of each cluster of YES supporters.

3. identifying_yes_no_supporters/ - The folder with data generated for the identitying yes and no supporters analysis part.

   3.1. identifying_yes_no_supporters/users_support.csv - Users and number of tweets produced by each one, by suport.

   3.2. identifying_yes_no_supporters/users_hits_yes.csv - Result of running HITS on YES supporters.

   3.3. identifying_yes_no_supporters/users_hits_no.csv - Result of running HITS on NO supporters.

   3.4. identifying_yes_no_supporters/kpp_score_yes.csv - Result of running KPP-NEG on yes supporters.

   3.5. identifying_yes_no_supporters/kpp_score_yes.csv - Result of running KPP-NEG on no supporters.

4. spread_of_influence/ - The folder with data generated for the spread of influence analysis part.

   4.1. spread_of_influence/spread_of_influence_lpa_m.csv - Result of running LPA on M users.

   4.2. spread_of_influence/spread_of_influence_lpa_m2.csv - Result of running LPA on M' users.

   4.3. spread_of_influence/spread_of_influence_lpa_k.csv - Result of running LPA on K users.

   4.4. spread_of_influence/spread_of_influence_modified_lpa_m.csv - Result of running modified LPA on M users.

   4.5. spread_of_influence/spread_of_influence_modified_lpa_2.csv - Result of running modified LPA on M' users.

   4.6. spread_of_influence/spread_of_influence_modified_lpa_k.csv - Result of running LPA on K users.

After running the generate_graphs script each folder will have the graphs used on the report under the figures folder.

6. Report

The report of the analysis with each step taken can be found under Report.pdf.