Lecture 2 – Types of Data and Patterns in Data Mining

Lecture 2 explains the types of data in Data Mining, including structured, semi-structured, unstructured, spatial, temporal, and multimedia data. It also covers data patterns, practical examples, case studies, diagrams, and tools used in industry perfect for BS CS, BS AI, BS IT & Data Science learners.

Data is the fuel of modern analytics, machine learning, artificial intelligence, and decision-making. But not all data is equal. Some data is neatly organized, while other data exists in raw, chaotic forms like tweets, images, logs, or sensor signals. Lecture 2 explores the types of data used in Data Mining and the patterns that help uncover meaningful insights.

Understanding the nature of data is essential before applying any mining algorithm. This lecture prepares you to handle complex datasets in real industrial environments such as healthcare, banking, retail, cybersecurity, and IoT.

Introduction to Data Types

Why Understanding Data Types Matters

If you choose the wrong method for the wrong data type, results become inaccurate.
For example:

• Classification works well with structured data
• NLP techniques are required for text data
• CNNs work best for images

How Data Types Influence Algorithms

Algorithms depend on the data structure.
Example:

• SQL queries → Structured data
• Deep learning → Unstructured data

Real-World Motivation

Every industry uses a mix of structured and unstructured data:

• Banks store transactional data
• Hospitals store medical records & X-rays
• Social media generates text, images, and videos
• IoT generates streaming sensor data

Structured Data

Structured data is organized in rows and columns, making it easy to search, analyze, and model.

Characteristics

• Tabular format
• Clearly defined schema
• Fast processing

Examples

• SQL databases
• Attendance sheets
• Inventory systems

Advantages

• Easy to manage
• Supports fast queries
• Ideal for statistical analysis

Limitations

• Cannot store video, images, or text efficiently
• Lacks flexibility for modern AI applications

Semi-Structured Data

Semi-structured data contains tags or markers but does not follow a strict schema.

XML & JSON

Used extensively in APIs, web applications, and configuration files.

Example (JSON):

{
  "name": "Ali",
  "age": 25,
  "skills": ["Python", "Data Mining"]
}

Log Files

• Server logs
• Application logs
• Security logs

Use Cases

• System monitoring
• Web analytics
• Big data pipelines

Lecture 1 – Introduction to Data Mining Concepts, Tasks & Real-World Applications

Unstructured Data

Unstructured data has no predefined model and is the largest type of data in the world.

Text Data

Tweets, comments, blogs, WhatsApp messages.

Techniques required:

• Tokenization
• Stemming
• Sentiment analysis
• Topic modeling

Multimedia Data

Images, video, and audio require:

• CNNs
• Speech recognition
• Face detection

Big Data Challenges

Unstructured data requires:

• High storage
• Specialized algorithms
• GPU computing

Specialized Data Types

Transactional Data

Each row represents a transaction.
Used in:

• Market basket analysis
• Fraud detection

Example:

Transaction ID | Items
001            | Milk, Bread, Eggs

AI Course → https://electuresai.com/artificial-intelligence

Temporal (Time-Series) Data

Data with time stamps:

• Stock prices
• Weather readings
• IoT sensor data

Spatial & Geographic Data

Used in:

• Google Maps
• GIS systems
• Disease mapping

Streaming Data

Continuous real-time data:

• Live temperature sensors
• Social media streams
• Network intrusion logs

Kaggle Learn → https://kaggle.com/learn

Data Patterns in Data Mining

Frequent Patterns

Patterns that appear frequently in datasets.

Example:

Customers who buy rice often buy oil.

Sequential Patterns

Patterns that follow a sequence.

Example:

Visit → Add to cart → Purchase

Graph Patterns

Used in:

• Social networks
• Knowledge graphs

Spatial Patterns

Used in:

• Earthquake analysis
• Weather prediction
• Location-based services

Data Characteristics That Affect Mining

Dimensionality

Higher dimensions = harder to analyze.
Example:
Image data has thousands of pixels (features).

Sparsity

Sparse datasets contain many zeros
Examples:

• Recommendation systems
• Text mining

Distribution & Noise

Noisy data reduces model accuracy.

Examples & Case Studies

Healthcare

• Patient vitals (temporal data)
• MRI images (unstructured data)
• Disease prediction (patterns)

Banking

• Credit score data (structured)
• Fraud detection (outlier detection)

E-Commerce

• User behavior (sequential patterns)
• Purchase history (transactional data)

Tools for Handling Different Data Types

Python Libraries

• Pandas → Structured
• NumPy → Numeric
• Matplotlib → Visualization
• NLTK, SpaCy → Text
• OpenCV → Image

NoSQL Databases

• MongoDB
• Cassandra
• Firebase

Cloud Storage

• AWS S3
• Google Cloud Storage
• Azure Blob

Common Mistakes When Handling Data Types

• Treating text as structured data
• Ignoring multimedia data
• Using wrong algorithms
• Not preprocessing properly
• Mixing inconsistent data formats

Summary

Lecture 2 introduced the essential types of data in Data Mining: structured, semi-structured, unstructured, temporal, spatial, multimedia, and transactional data. You explored how different data types require different tools, algorithms, and preprocessing steps. Finally, you learned how patterns such as frequent, sequential, graph, and spatial patterns help uncover hidden knowledge.

Next Lecture 3 – Data Preprocessing in Data Mining Cleaning, Transformation & Integration

People also ask: