Mining social network graphs

Next

Mining social network graphs

Iacob Isabela, BMDC

Social network graphs

2. Graph Presentation

1. Definition

3. Community detection

Back to Topics

Definition

What is mining social network graphs?

Mining social network graphs involves using graph theory, text mining, and machine learning techniques to analyze relationships and patterns within social networks to understand user behavior, detect communities, predict friendships, and identify influencers. Key applications include fraud detection, building recommendation systems, and targeted advertising

+ Info

Back to Topics

Graph Presentation

But how do we become graphs?

Social networks are commonly modeled as graphs, where each user is represented as a node, and their interactions or relationships are defined as edges. These edges can take different forms depending on the platform and type of interaction

+ Info

Back to Topics

Community Detection

What does it mean?

One of the main goals in social network analysis is to discover communities, groups of users who are more closely connected to each other than to the rest of the network. In simple terms, these are clusters of people who interact frequently, share common interests, or belong to the same social circle.

More

Go to map

Click for even more

Disjoint communities

• In disjoint (non-overlapping) communities, each user belongs to only one community.Ei mollis audire interpretaris cum.
• The communities are completely separated with no shared members.
• This is like dividing people into strict, non-mixing groups, for example, students being assigned to one class only (Class A or Class B).
• This is not very realistic for social networks.

Overlapping communities

• In overlapping communities, users can belong to multiple communities at the same time.
• This better reflects real life, because people often participate in multiple social circles (family, college friends, gaming group, coworkers, etc.).
• Much more complex to detect, but far more accurate for platforms like Facebook, Twitter, Reddit, or Discord.

What else?

Definition

Social Graph

Community Detection

Influence Discovery

Types of Community Detection

Recommenders

Anomaly Detection

End

2.Recommenders

1.. Influence discovery

3. Anomaly Detection

Influence Discovery

What are some other benefits?

Another important goal in social network mining is to identify key individuals who have a strong influence within the network. These are users who are highly connected, trusted, or strategically positioned, making them capable of spreading information quickly or shaping opinions.

Influencers

Hubs

Bridges

people with wide reach or strong impact (viral content creators)

users connected to many others

users who connect two otherwise separate communities

And more..

1. Influence discovery

2.Recommenders

3. Anomaly Detection

Recommenders

What are the recommendation systems?

Social network mining is often used to build recommendation systems that suggest relevant content, products, or new connections to users. These systems analyze both user behavior/preferences and the structure of the network to make intelligent predictions.

Posts, videos, or topics

New friends or followers

Products or ads

And more..

Recommenders

Influence discovery

AnomalyDetection

Anomaly Detection

An important application of social network mining is the detection of unusual or suspicious behavior that does not follow normal social patterns. This helps identify fraudulent, malicious, or automated activity, such as:

Repetitive or automated messaging behavior

Unusual connection density

Sudden behavior changes over time

Accounts that bridge unrelated communities unnaturally

The end..

What else?

Definition

Social Graph

Community Detection

Influence Discovery

Types of Community Detection

Recommenders

Anomaly Detection

End

Back to start

The end

Thank you!

Q&A

Weaknesses

Contextualize your topic

• Plan the structure of your communication.
• Give it a hierarchy and give visual weight to the main point.
• Add secondary messages with interactivity.
• Establish a flow through the content.
• Measure results.

But what are some types of interactions?

• Undirected edges are used when the relationship is mutual. For example, being "friends" or following each other on Facebook.
• Directed edges are used when the relationship has a direction, such as following someone on Instagram or Twitter.

• Weighted edges can include additional information, like how frequently two people interact, how strong their relationship is, or the recency of communication.

For example, messaging someone daily would have a higher weight than someone you contacted once six months ago.

Example: A simplified Facebook network where you are placed only in your family group OR only in your school group, but not both.

Example: A simplified Facebook network where you are placed only in your family group OR only in your school group, but not both.

Do we experience this?

Well, yes. This means predicting future connections between users. For example, when the app suggests "You might know Daniel" just because he’s friends with Emma, and Emma liked your cat picture once. The system basically assumes you two might get along because you're connected through mutual friends, kind of like a friendly nudge saying, "Hey, you’re all crazy about cats anyway."

Hierarchical Communities CommunitiesAre organized in layers, like a tree structure. A small group can be part of a larger group. Useful for analyzing large-scale networks where communities form natural sub-groups. Example: The main community forms "Music Fans" and a Sub-community of it would be "K-Pop Fans" Dynamic Communities These communities change over time. Users join, leave, or shift between communities. Used in scenarios like trend detection, evolving online group chats, or trending topics. Example: A temporary community forms around a viral hashtag, then fades as interest drops.

Threats

Contextualize your topic

• Plan the structure of your communication.
• Give it a hierarchy and give visual weight to the main point.
• Add secondary messages with interactivity.
• Establish a flow through the content.
• Measure results.