Lecture 1.

Cluster Analysis:
An Introduction
 Lecture 1. Cluster Analysis: An Introduction
 What Is Cluster Analysis?
 Applications of Cluster Analysis
 Cluster Analysis: Requirements and Challenges
 Cluster Analysis: A Multi-Dimensional Categorization
 An Overview of Typical Clustering Methodologies
 An Overview of Clustering Different Types of Data
 An Overview of User Insights and Clustering
 Summary
2
Session 1: What Is Cluster
Analysis?
 What Is Cluster Analysis?
 What is a cluster?
 A cluster is a collection of data objects which are
 Similar (or related) to one another within the same group (i.e., cluster)
 Dissimilar (or unrelated) to the objects in other groups (i.e., clusters)
 Cluster analysis (or clustering, data segmentation, …)
 Given a set of data points, partition them into a set of groups (i.e.,
clusters) which are as similar as possible
 Cluster analysis is unsupervised learning (i.e., no predefined classes)
 This contrasts with classification (i.e., supervised learning)
 Typical ways to use/apply cluster analysis
 As a stand-alone tool to get insight into data distribution, or
 As a preprocessing (or intermediate) step for other algorithms

4
Session 2: Applications of
Cluster Analysis
 Cluster Analysis: Applications
 A key intermediate step for other data mining tasks
 Generating a compact summary of data for classification, pattern discovery,
hypothesis generation and testing, etc.
 Outlier detection: Outliers—those “far away” from any cluster
 Data summarization, compression, and reduction
 Ex. Image processing: Vector quantization
 Collaborative filtering, recommendation systems, or customer segmentation
 Find like-minded users or similar products
 Dynamic trend detection
 Clustering stream data and detecting trends and patterns
 Multimedia data analysis, biological data analysis and social network analysis
 Ex. Clustering images or video/audio clips, gene/protein sequences, etc.
6
Session 3: Cluster Analysis:
Requirements and Challenges
 Considerations for Cluster Analysis
 Partitioning criteria

 Single level vs. hierarchical partitioning (often, multi-level hierarchical

 Exclusive (e.g., one customer belongs to only one region) vs. non-exclusive
(e.g., one document may belong to more than one class)
 Similarity measure

 Distance-based (e.g., Euclidean, road network, vector) vs. connectivity-

 Full space (often when low dimensional) vs. subspaces (often in high-
dimensional clustering)
8
 Requirements and Challenges
 Quality
 Ability to deal with different types of attributes: Numerical, categorical,
text, multimedia, networks, and mixture of multiple types
 Discovery of clusters with arbitrary shape
 Ability to deal with noisy data
 Scalability
 Clustering all the data instead of only on samples
 High dimensionality
 Incremental or stream clustering and insensitivity to input order
 Constraint-based clustering
 User-given preferences or constraints; domain knowledge; user queries
 Interpretability and usability

9
Session 4: Cluster Analysis: A
Multi-Dimensional Categorization
 Cluster Analysis: A Multi-Dimensional Categorization
 Technique-Centered
 Distance-based methods
 Density-based and grid-based methods
 Probabilistic and generative models
 Leveraging dimensionality reduction methods
 High-dimensional clustering
 Scalable techniques for cluster analysis
 Data Type-Centered
 Clustering numerical data, categorical data, text data, multimedia data, time-
series data, sequences, stream data, networked data, uncertain data
 Additional Insight-Centered
 Visual insights, semi-supervised, ensemble-based, validation-based
11
Session 5: An Overview of Typical
Clustering Methodologies
 Typical Clustering Methodologies (I)
 Distance-based methods
 Partitioning algorithms: K-Means, K-Medians, K-Medoids
 Hierarchical algorithms: Agglomerative vs. divisive methods
 Density-based and grid-based methods
 Density-based: Data space is explored at a high-level of granularity and
then post-processing to put together dense regions into an arbitrary shape
 Grid-based: Individual regions of the data space are formed into a grid-like
structure
 Probabilistic and generative models: Modeling data from a generative process
 Assume a specific form of the generative model (e.g., mixture of Gaussians)
 Model parameters are estimated with the Expectation-Maximization (EM)
algorithm (using the available dataset, for a maximum likelihood fit)
 Then estimate the generative probability of the underlying data points
13
 Typical Clustering Methodologies (II)
 High-dimensional clustering
 Subspace clustering: Find clusters on various subspaces
 Bottom-up, top-down, correlation-based methods vs. δ-cluster methods
 Dimensionality reduction: A vertical form (i.e., columns) of clustering
 Columns are clustered; may cluster rows and columns together (co-clustering)
 Probabilistic latent semantic indexing (PLSI) then LDA: Topic modeling of text data
 A cluster (i.e., topic) is associated with a set of words (i.e., dimensions) and a
set of documents (i.e., rows) simultaneously
 Nonnegative matrix factorization (NMF) (as one kind of co-clustering)
 A nonnegative matrix A (e.g., word frequencies in documents) can be
approximately factorized two non-negative low rank matrices U and V
 Spectral clustering: Use the spectrum of the similarity matrix of the data to
perform dimensionality reduction for clustering in fewer dimensions
14
Session 6: An Overview of
Clustering Different Types of Data
 Clustering Different Types of Data (I)
 Numerical data
 Most earliest clustering algorithms were designed for numerical data
 Categorical data (including binary data)
 Discrete data, no natural order (e.g., sex, race, zip-code, and market-basket)
 Text data: Popular in social media, Web, and social networks
 Features: High-dimensional, sparse, value corresponding to word frequencies
 Methods: Combination of k-means and agglomerative; topic modeling; co-clustering
 Multimedia data: Image, audio, video (e.g., on Flickr, YouTube)
 Multi-modal (often combined with text data)
 Contextual: Containing both behavioral and contextual attributes
 Images: Position of a pixel represents its context, value represents its behavior
 Video and music data: Temporal ordering of records represents its meaning
16
 Clustering Different Types of Data (II)
 Time-series data: Sensor data, stock markets, temporal tracking, forecasting, etc.
 Data are temporally dependent
 Time: contextual attribute; data value: behavioral attribute
 Correlation-based online analysis (e.g., online clustering of stock to find stock tickers)
 Shape-based offline analysis (e.g., cluster ECG based on overall shapes)
 Sequence data: Weblogs, biological sequences, system command sequences
 Contextual attribute: Placement (rather than time)
 Similarity functions: Hamming distance, edit distance, longest common subsequence
 Sequence clustering: Suffix tree; generative model (e.g., Hidden Markov Model)
 Stream data:
 Real-time, evolution and concept drift, single pass algorithm
 Create efficient intermediate representation, e.g., micro-clustering
17
 Clustering Different Types of Data (III)
 Graphs and homogeneous networks
 Every kind of data can be represented as a graph with similarity values as edges
 Methods: Generative models; combinatorial algorithms (graph cuts); spectral
methods; non-negative matrix factorization methods
 Heterogeneous networks
 A network consists of multiple typed nodes and edges (e.g., bibliographical data)
 Clustering different typed nodes/links together (e.g., NetClus)
 Uncertain data: Noise, approximate values, multiple possible values
 Incorporation of probabilistic information will improve the quality of clustering
 Big data: Model systems may store and process very big data (e.g., weblogs)
 Ex. Google’s MapReduce framework
 Use Map function to distribute the computation across different machines
 Use Reduce function to aggregate results obtained from the Map step
18
Session 7: An Overview of User
Insights and Clustering
 User Insights and Interactions in Clustering
 Visual insights: One picture is worth a thousand words
 Human eyes: High-speed processor linking with a rich knowledge-base
 A human can provide intuitive insights; HD-eye: visualizing HD clusters
 Semi-supervised insights: Passing user’s insights or intention to system
 User-seeding: A user provides a number of labeled examples, approximately
representing categories of interest
 Multi-view and ensemble-based insights
 Multi-view clustering: Multiple clusterings represent different perspectives
 Multiple clustering results can be ensembled to provide a more robust solution
 Validation-based insights: Evaluation of the quality of clusters generated
 May use case studies, specific measures, or pre-existing labels
20
Session 8: Summary
 Summary: Cluster Analysis—An Introduction
 What Is Cluster Analysis?
 Applications of Cluster Analysis
 Cluster Analysis: Requirements and Challenges
 Cluster Analysis: A Multi-Dimensional Categorization
 An Overview of Typical Clustering Methodologies
 An Overview of Clustering Different Types of Data
 An Overview of User Insights and Clustering
 Summary
22
 Recommended Readings
 Major Reference Books on Cluster Analysis

 Jiawei Han, Micheline Kamber, and Jian Pei. Data Mining: Concepts and Techniques.
Morgan Kaufmann, 3rd ed. , 2011 (Chapters 10 & 11)
 Charu Aggarwal and Chandran K. Reddy (eds.). Data Clustering: Algorithms and
Applications. CRC Press, 2014
 Mohammed J. Zaki and Wagner Meira, Jr.. Data Mining and Analysis: Fundamental
Concepts and Algorithms. Cambridge University Press, 2014
 Reference paper for this lecture

 Charu Aggarwal. An Introduction to Clustering Analysis. in Aggarwal and Reddy (eds.).

23