My FaceBook Networkwith Gephi

Coursera SNAAssignement

March, 18th 2013

Get my FaceBook graph Use getnet/netviz app to download my fb

network in gml format. Open it with Gephi.

Basic graph properties In the "context" panel (upper right), read: 712 nodes; 7433 edges; UNdirected graph (fb "friendship" is

symmetrical).

Layout adjustement Run a "Force Atlas 2" algorithm, setting

"Gravity" (under "tunings") to 25.0 (after a couple of explorative runs), to prevent a somehow large disconnected component to be pull off the graph.

Use Dissuade Hubs and Prevent Overlap.

The graph so far...

Set colors In Partition/nodes, refresh and then choose

"sex" in the cascade menu to color male and female differently.

Note: 49.58% female, 49.16% male, 1.26% undeclared.

Degree analysis Directed graph --> just one type of degree

(directed graphs have in-degrees and out-degrees)

In Statistics, run Average Degree Average degree expected to be #edges/#nodes =

7433 / 712 = 10.4. RESULT IS 20.8 (each edge counts twice)?

Degree analysis results There's a node with degree ~140 and one with

degree ~150 --> go to the DataLab to detect who they are...

Note that top-ranking nodes (by degree) are all males!

Adjusting size by degree Select Ranking/Nodes and Degree in the

cascade menu; Select the diamond icon (size/weight) and a

circular Spline (min = 1, max = 20).

The graph with degree ranking

Connected components As the graph is undirected, there are no strongly connected

components; In Statistics, run the Connected components algorithm: it

reveals 16 of them (0 to 15). In Partition / Nodes, refresh and select Component ID

Detecting the giant component The largest component comprehend 97% of the

nodes; To identify it, select Giant component in Filters /

Topology, drag it in the Queries area and finally Filter to obtain the result: there are 691 (out of 712) in the giant component.

Betweeness analysis In Statistics, run Avg. Path Length; The average path has length 3.67, the graph

diameter is 9 and its radius 5.

Rank by betweeness In Partition / Ranking, refresh and select Betweeness Centrality

(with size/weight selected); Use a circular spline, with min = 10 and max = 50; If necessary, run a Force Atlas layout with the Adjust by size box

checked and a very high (~500) Gravity.

Community detection Run Modularity in the Statistics panel (9

communities are identified); In Partition / Nodes, refresh and select

Modularity, then Apply.

Adding labels

Slide 1Slide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12Slide 13Slide 14Slide 15Slide 16