Scribd
Upload
9 views

Unsupervised Machine Learning for overview

Unsupervised learning is a machine learning approach that identifies patterns in unlabeled data, aiming to uncover hidden structures and relationships. Key types include clustering, dimensionality reduction, and anomaly detection, each with specific algorithms and applications such as customer segmentation and fraud detection. The process involves several steps from data collection to model evaluation, with performance metrics tailored for each type of analysis.

Uploaded by

ashafi.mscs19seecs
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
9 views

Unsupervised Machine Learning for overview

Unsupervised learning is a machine learning approach that identifies patterns in unlabeled data, aiming to uncover hidden structures and relationships. Key types include clustering, dimensionality reduction, and anomaly detection, each with specific algorithms and applications such as customer segmentation and fraud detection. The process involves several steps from data collection to model evaluation, with performance metrics tailored for each type of analysis.

Uploaded by

ashafi.mscs19seecs
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Unsupervised Machine Learning: Key Concepts & Algorithms

1. What is Unsupervised Learning?

 Definition: A type of machine learning where the model learns patterns from unlabeled
data.

 Goal: Find hidden structures, relationships, or groups in data.

 Examples: Customer segmentation, anomaly detection, topic modeling.

2. Types of Unsupervised Learning

A) Clustering (Grouping Similar Data Points)

 Objective: Divide data into meaningful clusters based on similarity.

 Examples: Customer segmentation, image segmentation, social network analysis.

 Common Algorithms:

o K-Means Clustering (Partitioning data into K clusters)

o Hierarchical Clustering (Building a tree-like structure of clusters)

o DBSCAN (Density-Based Clustering) (Finding clusters of different shapes and


handling noise)

o Gaussian Mixture Model (GMM) (Soft clustering using probability distributions)

o Spectral Clustering (Graph-based clustering for complex data)

B) Dimensionality Reduction (Reducing Data Complexity)

 Objective: Reduce the number of features while preserving important information.

 Examples: Image compression, feature selection before classification.

 Common Algorithms:

o Principal Component Analysis (PCA) (Transforms data into principal


components)

o t-SNE (t-Distributed Stochastic Neighbor Embedding) (Visualizing high-


dimensional data)
o Autoencoders (Deep Learning-based Feature Reduction)

C) Anomaly Detection (Detecting Unusual Patterns)

 Objective: Identify outliers or rare events in data.

 Examples: Fraud detection, network intrusion detection, manufacturing defect


detection.

 Common Algorithms:

o Isolation Forest (Randomly isolates anomalies)

o Local Outlier Factor (LOF) (Detects outliers based on density)

o One-Class SVM (Finds normal data patterns, marks anomalies)

3. Key Steps in Unsupervised Learning

1. Data Collection – Gather raw, unlabeled data.

2. Data Preprocessing – Normalize, scale, or clean data for analysis.

3. Feature Engineering – Select important variables or reduce dimensions.

4. Model Selection – Choose clustering, anomaly detection, or dimensionality reduction


techniques.

5. Model Training – Train the model to find hidden patterns.

6. Evaluation & Interpretation – Use metrics like silhouette score, inertia, or visualizations.

7. Real-World Application – Deploy insights for decision-making.

4. Performance Metrics for Unsupervised Learning

 For Clustering:

o Silhouette Score (Measures cluster separation)

o Davies-Bouldin Index (Evaluates clustering compactness & separation)

o Inertia (Within-cluster sum of squared distances, used in K-Means)

 For Dimensionality Reduction:


o Explained Variance Ratio (Measures how much information PCA retains)

o Reconstruction Error (Used in autoencoders to check data loss)

5. Supervised vs. Unsupervised Learning

Feature Supervised Learning Unsupervised Learning

Labeled Data Required Not Required

Goal Predict known outcomes Find hidden patterns

Algorithms Regression, Classification Clustering, Anomaly Detection

Examples Fraud detection, sentiment analysis Customer segmentation, topic modeling

6. Real-World Applications of Unsupervised Learning

✅ Customer Segmentation – Grouping customers for targeted marketing.


✅ Anomaly Detection – Identifying fraud or unusual transactions.
✅ Image Compression – Reducing image size while preserving key features.
✅ Recommendation Systems – Grouping users with similar preferences.
✅ Medical Diagnosis – Identifying unknown disease patterns.

You might also like