Network Reduction

Often times when working with big networks such as social networks there is too much data to be able to visualize it all at once or analyze. To make it easier to work with the data, we can either create a specific view of the data like when working with database tables, or we can use sampling to take samples of the data.

Views

To visualize the different views we will use the following network as an example, where the different colors represent different departments in a company:

• Blue: IT department
• Red: HR department
• Green: Management department

Local View

The local view focuses on a specific group of nodes and their connections. So the local view is a subset of the entire network that has been selected based on some criteria. For example if we have a network of people in a company, we can create a local view of the network that only contains people that are in the same department to analyze how they communicate with each other.

Below you can see the local view of the IT department:

Global View

The global view allows for a general view of the entire network. Here we summarize nodes to a single node based on some criteria. For example if we have a network of people in a company, we can create a global view of the network that summarizes all the people in the same department to a single node. This allows us to see how the different departments communicate with each other.

Below you can see the global view of the network (whether it is a good thing that HR an Management don't communicate directly is up for debate).

Context View

The context view is a combination of the local and global view. First we summarize the network to a global view. Then we pick a node in the global view and expand it again. If we have our previous example of a network of people in a company, we can create a context view of the network that summarizes all the people in the same department to a single node. Then we pick a department and expand it again to see how the people in that department communicate with the other departments.

From the graph below we could assume that Alice is the team lead of the IT department, since she is the one that talks to management.

Ego View

Ego/node/focus view is a view of the network that is centered around a specific node. In this view, the selected node is the "ego," and its immediate connections, i.e. its neighbours are analyzed.

Filtering Edges

Another common method is remove edges from the network based on some criteria. For example if we have a network of people in a company, we can remove all edges that are not between people in the same department to analyze how each department communicates within.

Or if we had a network with weights on the edges, we could remove all edges that have a weight below a certain threshold.

Inter and Intra-Edges

A form of filtering edges is to reduce a network down to its inter or intra-edges.

Inter-edges can be defined as the edges that connect vertices between two different groups or communities and intra-edges connect vertices within a group.

So if for example we have a graph containing people in a company, we can group them by their gender. Then we can for example only look at the edges between same gendered people (intra-edges) or between different gendered people (inter-edges).

Sampling

Sampling is the process of taking a subset of the data and working with that instead of the entire network.