Social Network Analysis

Social network analysis, usually referred to as SNA, creates a model and studies the relationships of a group of social entities that exist in the form of a network. An entity can be a person, a computer, or a web page, and a relationship can be a like, link, or friendship (that is, a connection between entities).

In this chapter, you'll learn about the following:

• Graphs, since social networks are usually represented in this form
• Important algorithms that are used to gain insights from a graph
• How to load, dump, and sample large graphs

Introduction to graph theory

Basically, a graph is a data structure that's able to represent relations in a collection of objects. Under this paradigm, the objects are the graph's nodes and the relations are the graph's links (or edges). The graph is directed if the links have an orientation (conceptually, they're like the one-way streets of a city); otherwise, the graph is undirected. In the following table, examples of well-known graphs are provided:

Graph example	Type	Nodes	Edges
World Wide Web	Directed	Web pages	Links
Facebook	Undirected	People	Friendship
Twitter	Directed	People	Follower
IP network	Undirected	Hosts	Wires/Connections
Navigation systems	Directed	Places/Addresses	Streets
Wikipedia	Directed	Pages	Anchor links
Scientific literature	Directed	Papers	Citations
Markov...

Graph algorithms

To get insights from graphs, many algorithms have been developed. In this chapter, we'll use a well-known graph in NetworkX, that is, the Krackhardt Kite graph. It is a dummy graph containing 10 nodes, and it is typically used to proof graph algorithms. David Krackhardt is the creator of the structure, which has the shape of a kite. It's composed of two different zones. In the first zone (composed of nodes 0 to 6), the nodes are interlinked; in the other zone (nodes 7 to 9), they are connected as a chain:

In: G = nx.krackhardt_kite_graph()
    nx.draw_networkx(G)
    plt.show()

In the following plot, you can examine the Krackhardt Kite's graph structure:

Let's start with connectivity. Two nodes of a graph are connected if there is at least a path (that is, a sequence of nodes) between them.

If at least one path exists, the shortest path between...

Graph loading, dumping, and sampling

Beyond NetworkX, graphs and networks can be generated and analyzed with other software. One of the best open source multiplatform software that can be used for their analysis is named Gephi. It's a visual tool and it doesn't require programming skills. It's freely available at http://gephi.github.io.

As in machine learning datasets, even graphs have standard formats for storing, loading, and exchanging. This way, you can create a graph with NetworkX, dump it to a file, and then load and analyze it with Gephi.

One of the most frequently used formats is Graph Modeling Language (GML). Now, let's see how we can dump a graph into a GML file:

In: dump_file_base = "dumped_graph"

    # Be sure the dump_file file doesn't exist
    def remove_file(filename):
        import os
        if os.path.exists(filename):
       ...

Summary

In this chapter, we learned what a social network is, including its creation and modification, representation, and some of the important measures of the social network and its nodes. Finally, we discussed the loading and saving of large graphs and ways to deal with the same.

With this chapter, almost all of the essential data science algorithms have been presented. Machine learning techniques were discussed in Chapter 4, Machine Learning, and social network analysis methods were discussed here. We will finally discuss the most advanced and cutting-edge techniques of deep learning and neural networks in the next chapter, Deep Learning Beyond the Basics.