UNT-II SOCIAL NETWORK ANALYSIS

Introduction to Social Networks Pro iles - De inition of SNA - Importance of SNA - Data
collection and preparation - Challenges and future Analysis of Social Networks Extracted from
Log Files - Data Mining Methods Related to SNA and Log Mining - Clustering Techniques -
Case Study.

Introduction to Social Network Profiles

Social networks are more than just apps on our phones; they are digital maps of human
relationships. At the core of these maps sits the profile, the fundamental unit that represents "you" in
the digital space.

In technical terms, a social network is a social structure made up of individuals (or organizations)
called "nodes," which are tied (connected) by one or more specific types of interdependency.

The Network Effect: The value of the network increases as more people join and create profiles.

The Goal: To facilitate communication, content sharing, and relationship building.

Profile in a Social Network:

A profile is a digital representation of an entity (person, business, or bot) within a platform. It serves
as your "ID card" and determines how the network’s algorithm interacts with you.

I. Identity: It defines who you are (Name, Bio, Photo).

II. History: It records your actions (Posts, Likes, Shares).

III. Gateway: It acts as the bridge that allows you to connect to others.

Attributes of a Profile (Nodes and Data)

In the study of Graph Theory—which powers social networks—we look at profiles through two
lenses: Nodes and Attributes.

• Node (Vertex)The individual point or "dot" in the network.A user, a brand page, or a group.
• AttributesThe specific data points attached to that [Link], location, interests, job title.
• Edges (Links)The connection between two nodes."Following," "Friendship," "Member of.”

Types of Social Network Profiles

Not all profiles serve the same purpose. They generally fall into three categories:

1. Personal Profiles: Focus on individual identity and private connections (e.g., a private
Instagram or Facebook account).
f
f
 2. Professional Profiles: Focus on career, skills, and industry networking (e.g., LinkedIn).

3. Public/Brand Profiles: Focus on reach, broadcasting, and engagement (e.g., a "Creator"

account on TikTok or a Business page on X/Twitter).

Definition of Social Network Analysis (SNA)

Social Network Analysis (SNA) is the "science of relationships." While a profile tells us who a
person is, SNA tells us how people are connected and how information flows between them. It is the
core methodology used by data scientists to understand everything from viral trends to the spread of
diseases.

Definition

Social Network Analysis (SNA) is the process of investigating social structures through the use of
networks and graph theory. It characterizes networked structures in terms of nodes (individual
actors, people, or things within the network) and the ties/edges (relationships or interactions) that
connect them.

Think of SNA as mapping and measuring the relationships between people or groups. Instead of
looking at individual data (like "this person lives in New York"), SNA looks at the "lines" between
people (like "this person is friends with 500 people in New York"). It shifts the focus from the
individual to the connection.

Key Components of SNA

To perform SNA, you must break a network down into three primary elements:

I. Nodes (Actors/Vertices): The entities in the network. These are usually people, but they can
also be organizations, websites, or even hashtags.

II. Edges (Ties/Links): The connections between nodes.

A. Directed Edges: One-way relationships (e.g., following someone on X/Twitter).

B. Undirected Edges: Mutual relationships (e.g., being "Friends" on Facebook).

III. Attributes: The specific properties of the nodes (e.g., age, job title) or the edges (e.g., the
"strength" of a friendship based on how often they message).

Relation Between Graph Theory and SNA

SNA is essentially applied Graph Theory. While Graph Theory is the mathematical study of points
and lines, SNA applies those math rules to human behavior.
 Concept in Graph Equivalent in
Purpose in Analysis
Theory SNA

Graph Social Network The entire structure being studied.

Vertex Actor / Pro le The individual "unit" of the network.
Edge Social Tie The interaction or relationship.
A mathematical way to record who is connected to
Adjacency Matrix Connection Table
whom.

• Degree Centrality: Who has the most connections? (The "Influencer").

• Betweenness Centrality: Who acts as a "bridge" between two different groups?

• Closeness Centrality: Who can spread information the fastest?

Fig 2.1: Simple node-edge graph

Reference Video link: [Link]

Importance of Social Network Analysis (SNA)

While standard statistics tell us what is happening (e.g., "10% of users are active"), Social Network
Analysis (SNA) tells us how it is happening by focusing on the relationships between people rather
than their individual characteristics.

Why SNA is Needed?

fi
Traditional data analysis often treats individuals as independent units. However, in the real world,
people influence each other. SNA is essential because:

• Shifts Focus to Context: It moves from analyzing attributes (age, gender) to analyzing
structures (who talks to whom).

• Reveals Hidden Patterns: It can identify "shadow" organizations—informal groups that

actually get work done regardless of the official hierarchy.

• Explains Contagion: It explains how ideas, behaviors, and viruses spread through "social
contagion."

Benefits of SNA

SNA provides a unique "X-ray" view of a community or organization:

• Identifying Key Actors: It pinpoints influencers, "gatekeepers" (who control info flow),
and "bridges" (who connect different groups).

• Detecting Communities: It automatically finds clusters or cliques within a large, messy

dataset.

• Optimizing Flow: By finding "bottlenecks," organizations can improve how information or

resources move through the network.

• Risk Management: It helps identify "isolated" individuals who might be at risk or "central"
nodes whose removal could collapse the network.

Areas Where SNA is Used

SNA is a cross-disciplinary tool with powerful applications in several fields:

Field Application

Business/Marketing Identifying "alpha consumers" to trigger viral marketing campaigns.

Tracking the spread of infectious diseases (Contact Tracing) and

Healthcare
analyzing doctor-patient in uence.

Mapping criminal or terrorist cells to identify and disrupt the

Law Enforcement
leadership.

Powering "Friend Suggestions" and determining which content goes

Social Media
viral on a feed.

Improving team collaboration by seeing which departments are

Organizational Behavior
working in "silos."
fl
Role of SNA in Decision Making

For leaders and managers, SNA acts as a Strategic Decision Support Tool:

1. Targeted Interventions: Instead of messaging everyone, leaders can target the "hubs"
(central nodes) to ensure 100% information coverage with 10% of the effort.

2. Managing Change: During a merger, SNA helps leaders see which groups are resisting
integration.

3. Crisis Response: In a network failure or data breach, SNA identifies which parts of the
system are most vulnerable.

4. Talent Management: It identifies "hidden experts"—people everyone goes to for advice

who might not have a high formal title.

Social Network Analysis plays a crucial role in analyzing connected data across multiple domains.
By focusing on relationships and interactions, SNA provides deep insights, improves decision-
making, and supports effective problem-solving in real-world applications.

Data Collection and Preparation

Data is the fuel for Social Network Analysis. However, raw data from platforms like X (Twitter),
LinkedIn, or Facebook is often "noisy" and unorganized. To perform meaningful analysis, a data
scientist must follow a rigorous pipeline to move from raw digital footprints to a structured
mathematical graph.

Data Collection and Preparation is a fundamental and core technical process in Social Network
Analysis (SNA). It involves gathering raw data from various sources and converting it into a clean,
structured, and usable form for network analysis. Since social network data is often large, complex,
and noisy, proper preparation is essential to ensure accurate analysis and meaningful results.

Sources of Social Network Data

Social network data can be obtained from different sources based on the domain of application.
Common sources include social media platforms where users interact through likes, comments,
shares, and follows. System-generated log files from websites, learning platforms, and enterprise
systems record user activities such as page visits and clicks. Communication data like emails, phone
records, and chat messages also form important network data. In some cases, surveys and
questionnaires are used to manually collect relationship data, especially in organizational or
academic environments. Open datasets and sensor-based systems further contribute to social
network data collection.

Eg: Social media platforms, Log files and system records, Communication and survey data, Open
and real-time data sources
Types of Data in Social Network Analysis

Social Network Analysis deals with different forms of data depending on how information is
generated and stored. Understanding these data types is essential to correctly extract nodes, edges,
and relationships.

I) Structured Data

Definition: Structured data is data that is well-organized in a fixed format, usually stored in tables,
databases, or spreadsheets. Each attribute has a predefined field, making the data easy to store,
search, and analyze.

Explanation: In SNA, structured data commonly represents nodes and edges in the form of edge
lists, adjacency matrices, or relational tables. Because of its clear structure, this data can be directly
used for network construction and analysis.

Example: A table containing columns such as User_ID, Friend_ID, Interaction_Count, and

Timestamp.
For example, a Facebook friendship list stored in a database is structured data.

II) Semi-Structured Data

Definition: Semi-structured data does not follow a rigid table format but still contains tags,
markers, or key-value pairs that provide some level of organization.
Explanation: In SNA, semi-structured data often comes from log files, JSON responses, XML
files, and API outputs. This data needs to be parsed and processed before it can be converted into
nodes and edges.

Example: A website log file recording user activities in JSON format such as page visits and clicks.
Social media API responses that return user interaction data in JSON format are semi-structured
data.

III) Unstructured Data

Definition: Unstructured data is data that does not have any predefined structure or format and is
difficult to analyze directly.

Explanation: In Social Network Analysis, unstructured data includes text posts, comments,
messages, images, audio, and videos. Relationships are not explicitly defined and must be extracted
using techniques such as text mining or content analysis.

Example: Twitter tweets, WhatsApp chat messages, email text content, or comments on social
media platforms are examples of unstructured data.

Data Collection Techniques in Social Network Analysis

Data collection techniques define how social network data is gathered from various sources. The
choice of technique depends on data availability, scale, and application domain.

I) API-Based Data Collection

Definition: API-based data collection uses official Application Programming Interfaces (APIs)
provided by platforms to access social network data in a controlled and authorized manner.

Explanation: APIs allow researchers to collect user profiles, interactions, and network
relationships while following platform policies. This method is reliable, secure, and commonly used
for large-scale data collection.

Example: Using the Twitter API to collect tweets, retweets, mentions, and follower relationships
for network analysis.

II) Web Scraping

Definition: Web scraping is the process of automatically extracting data from websites by parsing
HTML content.

Explanation: This technique is used when APIs are unavailable or limited. Scraping tools collect
visible data from web pages, but ethical and legal guidelines must be strictly followed.
Example: Extracting user comments and interaction counts from a public forum or blog website to
analyze user networks.

III) Log File Mining

Definition: Log file mining involves collecting and analyzing system-generated logs that record
user actions and events.

Explanation: Logs capture detailed interaction data such as logins, page visits, downloads, and
clicks, making them useful for constructing interaction networks.

Example: Analyzing learning management system logs to study student–student and student–
teacher interactions.

IV) Survey-Based Data Collection

Definition: Survey-based data collection gathers social network data using questionnaires or
interviews, where participants report their relationships or interactions.

Explanation: This technique is useful for small-scale or controlled studies where direct digital data
is unavailable.

Example: A classroom survey asking students to list their friends or study partners to construct a
social network.

V) Streaming Data Collection

Definition: Streaming data collection captures real-time, continuously generated data from social
systems.

Explanation: This technique is used for dynamic networks where relationships and interactions
change frequently over time.

Example: Collecting live chat messages from an online collaboration platform to analyze real-time
interaction networks.

Data Cleaning

Raw social network data often contains errors such as missing values, duplicate records, noise, and
irrelevant interactions. Data cleaning focuses on removing duplicate nodes and edges, correcting
inconsistent values, handling missing data, and filtering spam or bot-generated interactions. This
step ensures that the data accurately represents real-world social interactions and prevents
misleading analysis results.

Actions: Remove duplicates and noise, Handle missing and incorrect data, Filter spam and
irrelevant interactions
Data Transformation

After cleaning, the data is transformed into a format suitable for network analysis. This involves
converting raw data into nodes and edges, assigning weights, directions, and labels to connections,
and aggregating multiple interactions into meaningful relationships. Data may also be converted
into adjacency matrices or edge lists depending on the analysis method.

Actions: Convert data into nodes and edges, Assign weights and directions, Prepare network
structures

Data Preprocessing Steps

Data preprocessing prepares the transformed data for efficient analysis. This includes filtering
inactive nodes, setting thresholds for edges, segmenting data based on time, and extracting
important features such as interaction frequency or centrality values. Normalization and validation
are performed to ensure consistency and accuracy before applying SNA algorithms.

Actions: Node and edge filtering, Feature extraction, Normalization and validation

Data Collection and Preparation is the backbone of Social Network Analysis. Proper handling of
social network data ensures reliable analysis, accurate insights, and effective decision-making.
Without systematic preparation, SNA techniques cannot produce meaningful results.

Challenges of Social Network Analysis

1. Social network data often contains sensitive personal information. Unauthorized access or
misuse of such data can violate privacy. Strong security and privacy-preserving methods are
required.
 2. Social networks are very large and complex in size. Processing millions of nodes and
connections is computationally difficult. Scalable algorithms and distributed systems are
needed.

3. Social networks continuously change over time. Users may join, leave, or alter their
interactions frequently. Capturing these dynamic changes is technically challenging.

4. Data collected from social platforms may be incomplete or noisy. Spam accounts and fake
interactions reduce data quality. Effective data cleaning is necessary for accurate analysis.

5. Ethical concerns arise when data is collected without consent. Analysis results may be
biased or misused. Responsible and transparent usage of data is essential.

6. Integrating data from multiple platforms is difficult. Different formats and standards cause
compatibility issues. Data integration requires additional preprocessing efforts.

7. Visualization of large social networks is complex. Dense networks are difficult to interpret
visually. Advanced visualization techniques are required.

Future Analysis of Social Networks

1. Artificial intelligence will play a major role in SNA. Machine learning models will improve
pattern detection. Prediction accuracy will significantly increase.

2. Real-time social network analysis will become more common. Live data streams will be
analyzed instantly. This is useful for security and trend monitoring.

3. Dynamic network analysis will receive more focus. Time-aware models will track network
evolution. This will improve understanding of behavior changes.

4. Privacy-preserving analysis techniques will advance. Anonymization and secure

computation will be improved. User trust and legal compliance will increase.

5. Multilayer and heterogeneous network analysis will expand. Multiple types of relationships
will be studied together. This provides richer and more realistic insights.

6. Explainable social network models will be developed. Results will be easier for humans to
understand. Decision-making will become more transparent.

7. Cross-domain and interdisciplinary research will grow. Social networks will be combined
with healthcare, finance, and education. This will broaden the applications of SNA.

Future Trends and Research Directions

Trend Description

Using Large Language Models (LLMs) to understand not just who is

AI-Driven SNA
connected, but the nuanceand intent of their conversations.
Hyper-Personalized Social networks acting as primary search engines, using SNA to provide
Discovery answers based on "trusted" contacts rather than generic web results.

Multi-Layer Analyzing a person across multiple platforms simultaneously (e.g., how a

(Multiplex) Networks person’s LinkedIn professional ties in uence their TikTok content).

Visualizing 3D social networks in the Metaverse to understand spatial

AR/VR Integration
relationships and group dynamics.

Roadmap for Researchers

To stay relevant, future analysts must focus on:

1. Privacy-Preserving Analysis: Developing algorithms that can analyze a network without

ever seeing the raw, private data (Differential Privacy).

2. Hybrid Methods: Combining quantitative SNA (numbers) with qualitative "Social

Listening" (empathy and context).

3. Real-Time Intervention: Using SNA to stop "fake news" or "online toxicity" at the
moment it starts to spread from a central node.

Social Networks Extracted from Log Files

In modern information systems, user interactions are continuously recorded in the form of log files.
These log files provide a rich source of data for extracting social networks. By analyzing logs,
implicit relationships between users, systems, or entities can be identified. Social networks
extracted from log files play a crucial role in understanding user behavior, system usage patterns,
and interaction structures, making them an important link between Social Network Analysis (SNA)
and Data Mining.

Defintion:

Log files are automatically generated records that capture events occurring within a system or
application. Each log entry typically contains information such as user ID, timestamp, event type,
and system resource accessed. Logs are created by web servers, applications, databases, learning
management systems, and network devices. Since logs record real user actions, they provide
accurate and unbiased data for social network extraction.
fl
Types of Log Files

A) Web Server Logs

Logs that record user requests made to a web server. They capture page visits, clicks, and browsing
paths. Example: Apache or Nginx access logs.

B) Application Logs

Logs generated by software applications. They store user actions like login, upload, or form
submission. Example: Logs from an online shopping website.

C) System Logs

Logs generated by the operating system. They record system events such as startup and errors.
Example: Linux system event logs.

D) Security Logs

Logs that track authentication and access control events. Used to detect unauthorized access or
attacks. Example: Login failure and firewall logs.

Process of Extracting Networks from Log Files

The extraction process begins with collecting raw log files from relevant systems. These logs are
then cleaned to remove noise, duplicate entries, and irrelevant events. Next, important entities such
as users, sessions, or resources are identified as nodes. Relationships between these entities are
derived based on shared activities, sequences of events, or interaction frequency, forming edges in
the network. Finally, the extracted network is represented as a graph and prepared for analysis using
SNA techniques.
User Behavior Analysis Using Log-Based Networks

Log-based social networks help analyze user behavior by revealing interaction patterns and usage
trends. By studying connections derived from logs, analysts can identify active users, collaboration
patterns, and frequently accessed resources. This analysis is useful in understanding learning
behavior in educational systems, customer behavior in e-commerce platforms, and employee
activity in organizational systems. Log-based networks provide insights into both individual and
group-level behavior.

Event-Based Networks

Event-based networks are constructed by treating events recorded in logs as the basis for
relationships. In this approach, nodes represent users or system components, while edges represent
shared or sequential events. For example, two users accessing the same resource within a short time
interval may be connected. Event-based networks are especially useful for analyzing temporal
behavior, detecting anomalies, and understanding workflows within systems.

Importance in Data Mining and SNA

Extracting social networks from log files enables the application of data mining techniques such as
clustering, classification, and pattern discovery. These networks allow analysts to uncover hidden
relationships, predict future behavior, and optimize system performance. This approach strengthens
the integration of Social Network Analysis with log mining and practical real-world applications.

Social networks extracted from log files provide a powerful and practical method to study
interactions in digital environments. By converting raw log data into meaningful network structures,
organizations can gain valuable insights into user behavior, system usage, and interaction dynamics.
This makes log-based social network extraction a vital component of modern SNA and data mining
systems. Social networks extracted from log files transform raw system data into meaningful
interaction graphs. This approach supports behavioral analysis, system optimization, and intelligent
decision-making, making it a vital component of modern Social Network Analysis.

Challenges in Log Extraction

• Data Volume: Logs grow by gigabytes per hour; parsing them requires high-performance
tools (like Grep, AWK, or Python’s Pandas).

• Dynamic Identity: One person might use three different devices (IP addresses), making it
hard to realize they are the same "node."

• Privacy: Logs often contain sensitive metadata (location, time of activity) that must be
handled with extreme care.
Data Mining Methods Related to SNA and Log Mining
Data mining methods play a crucial role in extracting meaningful knowledge from social networks
and log data. Once interaction networks are built from log files, data mining techniques are applied
to analyze patterns, behaviors, and irregularities. These methods help convert raw interaction data
into actionable insights.

i) Classification Methods

Definition:
Classification is a supervised data mining technique that assigns nodes or data instances into
predefined categories based on learned patterns. In Social Network Analysis, classification is used
to categorize users according to their interaction behavior, role, or importance in the network.
Features such as degree centrality, number of interactions, session duration, and access frequency
derived from log files are used as input. A trained model predicts the class of unknown users or
events.

SNA & Log Mining Relevance:

Classification helps in identifying user roles, predicting behavior, and managing large networks
efficiently.

Example:
Classifying users as normal, power users, or suspicious users based on their activity patterns in
web server logs.

ii) Association Rule Mining

Definition:
Association rule mining discovers frequent relationships and co-occurrence patterns among data
items. In SNA and log mining, association rules help identify relationships between users, actions,
or resources that often occur together. These rules are evaluated using measures such as support and
confidence. This method reveals hidden dependencies within interaction data.

SNA & Log Mining Relevance:

It helps in understanding interaction behavior, recommendation generation, and relationship
discovery in networks.

Example:
Finding that users who frequently access a forum also tend to download related documents from
system logs.

iii) Pattern Discovery

Definition:
Pattern discovery identifies recurring sequences or structures in data. In Social Network Analysis,
pattern discovery is used to detect frequent interaction paths, repeated communication structures, or
common behavioral sequences. Log data provides time-ordered events that help identify these
repeated patterns. Such patterns represent normal or expected behavior within the network.

SNA & Log Mining Relevance:

This method supports behavior modeling and system optimization.

Example:
Discovering a repeated pattern where users log in → view content → participate in discussions →
log out.

iv) Anomaly Detection

Definition:
Anomaly detection identifies behaviors or data points that deviate significantly from normal
patterns. In SNA and log mining, anomalies may indicate security threats, fraud, or abnormal user
behavior. Network-based anomalies include unusually high connectivity, sudden spikes in activity,
or unexpected interaction paths. This method compares current behavior with normal network
patterns.

SNA & Log Mining Relevance:

Anomaly detection is essential for cybersecurity, fraud detection, and system monitoring.

Example:
Detecting a user account that suddenly communicates with many users in a short period, indicating
possible misuse.

Relationship Between Methods and SNA

These data mining methods enhance Social Network Analysis by adding intelligence to network
structures extracted from log files. Classification predicts roles, association rules reveal
relationships, pattern discovery identifies behavior trends, and anomaly detection ensures security
and reliability. Together, they form a powerful framework for analyzing social networks derived
from log data.

Social Network Analysis

Aspect Log Mining Data Mining
(SNA)
Extraction of useful Analysis of relationships Discovery of patterns
Basic Meaning
information from log les and interactions and knowledge from data

Structured, semi-
Type of Data Raw system and user Nodes and edges
structured, and network
Used activity logs representing interactions
data

Patterns, trends,
Events, actions, and Connections, in uence,
Main Focus prediction, and
timestamps and network structure
anomalies
fl
fi
 Role in Provides raw interaction Converts data into Analyzes networks to
Work ow data network graphs extract insights

Cleaned and processed log Rules, patterns, classes,

Output Social network graph
events and anomalies

Fraud detection, behavior

Example Web server access logs User interaction network
prediction
Intelligent decision
Importance Foundation for analysis Structural understanding
making

Clustering Techniques in Social Network

Analysis
Clustering is a fundamental technique in data mining and Social Network Analysis (SNA) used to
group similar entities together. In social networks, clustering helps identify communities, groups,
and functional structures by analyzing interaction patterns. Since social networks are large and
complex, clustering plays a crucial role in simplifying network analysis and discovering hidden
relationships.

Definition

Clustering is an unsupervised learning technique that groups data points or nodes into clusters
such that nodes within the same cluster are more similar to each other than to nodes in other
clusters. Unlike classification, clustering does not require predefined labels. In SNA, similarity is
often measured using interaction frequency, connectivity, or structural properties. Clustering helps
in discovering natural groupings within social networks.

Example: Grouping users in a social media platform based on interaction frequency

and shared interests.

Role of Clustering in Social Network Analysis

Clustering plays a vital role in understanding social network structures. It helps identify
communities where users interact more frequently with each other than with outsiders. Clustering
simplifies large networks by breaking them into manageable sub-networks. It supports applications
such as influence analysis, recommendation systems, fraud detection, and behavior prediction. In
log-based networks, clustering helps group users with similar activity patterns.
fl
Types of Clustering Techniques

Clustering techniques used in SNA can be broadly categorized based on how clusters are formed
and represented.

I) Partition-Based Clustering

Definition:
Partition-based clustering divides data into a fixed number of clusters. Each node belongs to exactly
one cluster. These methods aim to minimize intra-cluster distance and maximize inter-cluster
distance. Common Algorithm: K-Means

Example:
Grouping users into K clusters based on interaction frequency.

II) Hierarchical Clustering

Definition:
Hierarchical clustering builds a hierarchy of clusters.
It can be agglomerative (bottom-up) or divisive (top-down). It is useful for understanding multi-
level community structures.

Example:
Organizing employees into teams, departments, and divisions.

III) Density-Based Clustering

Definition:
Density-based clustering groups nodes based on dense regions.
Clusters are formed where nodes have high local density. It can detect noise and outliers. Common
Algorithm: DBSCAN

Example:
Detecting tightly connected user groups and isolated users.

IV) Graph-Based Clustering

Definition:
Graph-based clustering uses graph properties to detect clusters. It directly analyzes network
topology using edges and connectivity. Common Algorithms: Spectral clustering, Min-cut
methods

Example:
Partitioning a network into strongly connected components.
Community Detection

Definition

Community detection is a special form of clustering in SNA that identifies groups of nodes with
dense internal connections and sparse external connections. Communities represent real-world
social groups such as friend circles, work teams, or interest groups. Community detection
algorithms optimize measures like modularity to identify these structures.

Common Algorithms

• Girvan–Newman Algorithm

• Louvain Algorithm

• Label Propagation

Example

Detecting friend communities in a social media network.

Cluster Network diagram

Comparison of Clustering Techniques
Technique Key Idea Advantages Limitations
Distance-based
K-Means Simple and fast Requires K value
partitioning
Computationally
Hierarchical Tree-based clustering Multi-level insight
expensive

DBSCAN Density-based Handles noise well Parameter sensitive

Graph-Based Uses graph structure Suitable for networks Complex computation

Community Realistic
Modularity optimization Scalability issues
Detection communities

Clustering techniques are essential tools in Social Network Analysis for discovering communities
and simplifying complex networks. Different clustering methods serve different purposes depending
on network size, density, and application. Understanding these techniques is crucial for analyzing
real-world social networks and log-based interaction data.

Case Study: Log-Based Social Network

Analysis of Student Interaction in an E-
Learning Platform

Introduction
With the rapid growth of digital education platforms, understanding how students interact within
online learning environments has become essential. Traditional performance evaluation focuses
mainly on grades and attendance, but these measures fail to capture the underlying interaction
patterns among learners. Social Network Analysis (SNA), combined with log mining techniques,
provides a powerful approach to analyze student collaboration, engagement, and learning behavior
using system-generated data.

This case study presents a comprehensive analysis of student interaction networks extracted from
log files of an e-learning system.

Problem Description
The institution observed several challenges in its online learning platform: Uneven student
participation in discussion forums, Lack of collaboration among students, Poor engagement from a
subset of learners, Difficulty in identifying influential and isolated students
The main problem was the absence of a structured method to analyze interaction patterns among
students. The goal of this case study is to apply Social Network Analysis on log data to:

• Identify student communities

• Detect influential and isolated students

• Understand behavioral patterns

• Improve learning engagement strategies

Data Used

Data Source

The dataset was collected from the Learning Management System (LMS) of a university over one
academic semester (6 months). The data was extracted from system log files automatically
generated by the LMS.

Data Attributes

Each log entry contained the following information: Student ID, Session ID, Timestamp, Type of
activity (login, forum post, reply, resource access), Resource ID

Nature of Data

• Type: Semi-structured log data

• Volume: Large-scale (thousands of log entries)

• Dynamic: User interactions changed over time

• Sensitive: Contained personal student information

Data Collection and Preprocessing

Data Cleaning

Raw log data contained duplicate entries, missing values, and inactive users. The following steps
were performed:

• Removal of duplicate log records

• Filtering inactive students with negligible activity

• Correction of timestamp inconsistencies

• Removal of irrelevant system-generated events

Data Transformation

The cleaned data was transformed into a network-friendly format:

• Each student was considered a node

• Each interaction (reply, message, shared resource) was considered an edge

• Edge weights represented interaction frequency

Network Construction

A social network graph was constructed where:

• Nodes represented students

• Directed edges represented interaction flow

• Weighted edges reflected interaction strength

Method Applied

Social Network Analysis Techniques

The following SNA techniques were applied:

• Degree Centrality: To identify active students

• Betweenness Centrality: To identify influencers and mediators

• Clustering Coefficient: To measure collaboration strength

Clustering and Community Detection

Community detection algorithms were applied to identify student groups:

• Graph-based clustering to detect tightly connected groups

• Modularity optimization to evaluate community strength

Log Mining Integration

Data mining techniques such as:

• Pattern discovery to identify repeated interaction sequences

• Classification to categorize students as active, moderate, or inactive

Results

The analysis produced the following results:

 • Multiple distinct student communities were identified

• A small group of students exhibited high centrality and acted as knowledge hubs

• Several students were found to be isolated with minimal interaction

• Communities with strong internal connections showed higher academic performance

• Interaction frequency was directly linked to engagement levels

Interpretation of Results

The results indicate that student interaction strongly influences learning outcomes. Influential
students played a key role in knowledge dissemination and peer support. Isolated students were at
risk of disengagement and required targeted intervention. The network structure revealed that
collaborative learning improves overall academic performance.

Advantages of Using SNA in This Case

• Uses real-time, unbiased log data

• Captures hidden interaction patterns

• Scalable to large student populations

• Integrates well with data mining techniques

Limitations

• Privacy concerns regarding student data

• Dynamic nature of interactions requires continuous analysis

• Results depend on data quality and preprocessing accuracy

Conclusion

This case study clearly demonstrates that Social Network Analysis applied to log files is an
effective method for understanding student interaction behavior in e-learning environments. By
transforming raw log data into meaningful network structures, educators can gain valuable insights
to enhance learning engagement and academic success.

Other Case Study References:

1. Social Network Analysis of Online Customer Behavior Using E Commerce Logs

Ramesh, K., & Srinivasan, P. (2024). Social Network Analysis of Online Customer Behavior
Using E Commerce Logs. International Journal of Data Mining and Analytics, 12(3), 145–
162.
‑
‑
2. Employee Collaboration Network Analysis Using Enterprise System Logs
Sharma, A., Gupta, L., & Menon, S. (2023). Employee Collaboration Network Analysis
Using Enterprise System Logs. Journal of Organizational Computing and Electronic
Commerce, 29(4), 299–319.

3. Cybersecurity Threat Detection Using Log Based Social Networks

Bhattacharya, M., & Desai, R. (2025). Cybersecurity Threat Detection Using Log Based
Social Networks. IEEE Transactions on Network Science and Engineering, 11(1), 78–92.

4. Influence Spread Analysis in Social Media Platforms

Lee, J., & Kaur, H. (2024). Influence Spread Analysis in Social Media Platforms. Social
Network Analysis and Mining, 14(2), 87–105.

5. Healthcare Referral Network Analysis Using Patient Interaction Data

Patel, D., Verma, S., & Rao, P. (2023). Healthcare Referral Network Analysis Using Patient
Interaction [Link] of Healthcare Informatics Research, 7(1), 50–74.

6. Knowledge Sharing Network Analysis in Research Organizations

Singh, R., & Chatterjee, T. (2025). Knowledge Sharing Network Analysis in Research
[Link] Review, 18(5), 250–275.
‑
‑