Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 8 views

    ISLR Chap 10 Shaheryar-Mutahira

    Uploaded by

    Shaheryar Zahur
    This document provides an overview of unsupervised learning techniques including principal component analysis (PCA) and clustering. It describes PCA as an unsupervised approach to data visualization that identifies principal components as linear combinations of variables with the largest variance. It also outlines K-means and hierarchical clustering as methods to find homogeneous subgroups within data, with K-means partitioning observations into a pre-specified number of clusters and hierarchical clustering using a tree-based representation. The document discusses practical considerations for these unsupervised learning algorithms.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd

    ISLR Chap 10 Shaheryar-Mutahira

    Uploaded by

    Shaheryar Zahur
    8 views17 pages
    This document provides an overview of unsupervised learning techniques including principal component analysis (PCA) and clustering. It describes PCA as an unsupervised approach to data visualization that identifies principal components as linear combinations of variables with the largest variance. It also outlines K-means and hierarchical clustering as methods to find homogeneous subgroups within data, with K-means partitioning observations into a pre-specified number of clusters and hierarchical clustering using a tree-based representation. The document discusses practical considerations for these unsupervised learning algorithms.

    Copyright

    © © All Rights Reserved

    Available Formats

    PPTX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document provides an overview of unsupervised learning techniques including principal component analysis (PCA) and clustering. It describes PCA as an unsupervised approach to data visualization that identifies principal components as linear combinations of variables with the largest variance. It also outlines K-means and hierarchical clustering as methods to find homogeneous subgroups within data, with K-means partitioning observations into a pre-specified number of clusters and hierarchical clustering using a tree-based representation. The document discusses practical considerations for these unsupervised learning algorithms.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    8 views17 pages

    ISLR Chap 10 Shaheryar-Mutahira

    Uploaded by

    Shaheryar Zahur
    This document provides an overview of unsupervised learning techniques including principal component analysis (PCA) and clustering. It describes PCA as an unsupervised approach to data visualization that identifies principal components as linear combinations of variables with the largest variance. It also outlines K-means and hierarchical clustering as methods to find homogeneous subgroups within data, with K-means partitioning observations into a pre-specified number of clusters and hierarchical clustering using a tree-based representation. The document discusses practical considerations for these unsupervised learning algorithms.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    of 17

    Unsupervised Learning

    Mutahira Khalid Chaudry & Syed Shaheryar Zahur

    Convergent Business Technologies.


    Overview
    • Approach
    • Principal Component Analysis
    • Clustering

    Convergent Business Technologies.


    Unsupervised Learning
    • Definition
    • Goal in unsupervised learning
    • More challenging than supervised

    Convergent Business Technologies.


    Principal Component Analysis
    • Definition
    • An unsupervised approach
    • Serves as a tool for data visualization

    Convergent Business Technologies.


    Principal Components
    • Definition
    • Mathematical form

    • Loading vectors define direction

    • Squared sum of loading vectors should equal to 1

    Convergent Business Technologies.


    Principal Components
    • Variables centered to mean 0
    • Linear combination with largest variance

    • Second PC is orthogonal to PC1


    • PC1 and PC2 are uncorrelated

    Convergent Business Technologies.


    Example
    • PC1 weights on crimes
    • PC2 weights on UrbanPop
    • States with large positive score
    have high crime rates

    Convergent Business Technologies.


    More on PCA
    • Scaling variables
    • Uniqueness of PC
    • Proportion of variance
    • Deciding on number of PC

    Scree Plot

    Convergent Business Technologies.


    Clustering Methods
    • Techniques for finding sub-groups/clusters
    • Find homogenous subgroups among observations

     Two approaches
    I. K-means Clustering:
    partition observations into pre-specified number of clusters

    II. Hierarchical Clustering:


    Tree-like visualization of observations - dendrogram

    Convergent Business Technologies.


    K-means Clustering
    • Simple and elegant
    • Partition into distinct, non-overlapping clusters
    • Number of clusters specified beforehand

    Convergent Business Technologies.


    K-means Clustering

    Convergent Business Technologies.


    K-means Clustering

    Objective value:

    Convergent Business Technologies.


    Hierarchical Clustering
    • Does not require choice of K
    • Attractive tree-based representation - dendrogram
    • Bottom-up/Agglomerative clustering
    • Built starting from leaves
    • Combining clusters up to trunk

    Convergent Business Technologies.


    Hierarchical Clustering
    • Obtaining number of clusters

    Convergent Business Technologies.


    Hierarchical Clustering
    • Choice of dissimilarity important:
    • Linkage defines dissimilarity
    I. Complete Linkage
    II. Single Linkage
    III. Average Linkage
    IV. Centroid Linkage

    Convergent Business Technologies.


    Practical Issues in Clustering
    • Consideration of dissimilarity measure (hierarchical clustering)
    • Consideration of standardization (both approaches)
    • Not very robust models

    Convergent Business Technologies.


    Thank you.

    ©2019 Convergent Business Technologies

    Convergent Business Technologies.

    You might also like