Your first task is create a weighted directed graph with igraph,1 using raw data from Twitter. You will download a compressed file with Tweets posted during July 2009. The format of the file is the following:

T 2009-07-01 00:04:20 U http:https://twitter.com/greeneyed_panda W @Dprinzessin jajajajajajaja.......... no.....

In the tweet above, T indicates the time the tweet was posted (2009-07-01 00:04:20), U indicates the user that posted it (greeneyed panda) and W is the text of the tweet (@Dprinzessin jajajajajajaja.......... no.....). Twitter subscribers can use the @ character to make a mention to another subscriber, e.g. in the above tweet, user greeneyed panda made a mention to user Dprinzessin.

You will first manipulate the raw data with the programming language of your choice to create a total of 5 .csv files, one for each of the first five days of July 2009, using the following format:

from, to, weight user1, user2, 5 user2, user1, 1 user1, user3, 2 ...

Each .csv file should describe the weighted directed mention graph for the respective day, e.g., in the example above user1 has made 5 mentions to user2, user2 has made 1 mention to user1, and user1 has made 2 mentions to user3. Having created the .csv files it will be trivial to use them and create the respective igraph graphs.

Your submission should include the code you used to create the .csv files (any programming language), the code you used to create the igraph graphs (R) and the 5 (compressed) .csv files.

Your next task is to create plots that visualize the 5-day evolution of different metrics for the graph. More speci cally, you will create plots for:

1. Number of vertices
2. Number of edges
3. Diameter of the graph
4. Average in-degree
5. Average out-degree

What do you notice for each of the 5 above metrics? Are there significant fluctuations during these five days?

Next, you will write to code to create and print data frames for the 5-day evolution of the top-10 Twitter users with regard to:

1. In-degree
2. Out-degree
3. PageRank

Again, provide short comments on your findings. Do you notice variations on the top-10 lists for the different days?

Your final task is to perform community detection on the mention graphs. Try applying fast greedy clustering, infomap clustering, and louvain clustering on the undirected versions of the 5 mention graphs. Are you able to get results with all methods? Include a short comment on your report regarding the performance of the 3 algorithms.

Then, pick one of the three methods as well as a random user that appears in all 5 graphs and write code to detect the evolution of the communities this user belongs to. Do you spot similarities in the communities?

Finally, you will create a visualization of the graph using a different color for each community. Make sure to have a look at the sizes of the communities and filter out all nodes that belong to very small or very large communities, in order to create a meaningful and aesthetically pleasing visualization.