Scribd Logo
Upload

Welcome to Scribd!

  • 8 views

    10c UnsupervisedMethods

    Uploaded by

    LEM tv
    Hierarchical Clustering

    Copyright:

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

    10c UnsupervisedMethods

    Uploaded by

    LEM tv
    8 views5 pages
    Hierarchical Clustering

    Original Title

    10c_UnsupervisedMethods

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Hierarchical Clustering

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    8 views5 pages

    10c UnsupervisedMethods

    Uploaded by

    LEM tv
    Hierarchical Clustering

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 5

    Changelog:

     12  Dec  2016  


     
     
    Advantages  &  Disadvantages  of    
    k-­‐Means  and  Hierarchical  clustering  
    (Unsupervised  Learning)  
     
     
    Machine  Learning  for  Language  Technology  
    ML4LT  (2016)  
    Marina  San(ni  
    Department  of  LinguisHcs  and  Philology  
    Uppsala  University  
     
    2016    
    Advantages  &  Disadvantages  of  k-­‐Means  and  Hierarchical  Clustering   1  
    Outline
    •  k-­‐Means:  Advantages  and  Disadvantages  
    •  Hierarchical  Clustering:  Advantages  and  Disadvantages  

    2016   Advantages  &  Disadvantages  of  k-­‐Means  and  Hierarchical  Clustering   2  


    k-­‐Means:  Advantages  and  Disadvantages  
    Advantages  
    •  Easy  to  implement  
    •  With  a  large  number  of  variables,  K-­‐Means  may  be  computaHonally  faster  than  
    hierarchical  clustering  (if  K  is  small).  
    •  k-­‐Means  may  produce  Hghter  clusters  than  hierarchical  clustering  
    •  An  instance  can  change  cluster  (move  to  another  cluster)  when  the  centroids  are  re-­‐
    computed.    
    Disavantages  
    •  Difficult  to  predict  the  number  of  clusters  (K-­‐Value)  
    •  IniHal  seeds  have  a  strong  impact  on  the  final  results  
    •  The  order  of  the  data  has  an  impact  on  the  final  results  
    •  SensiHve  to  scale:  rescaling  your  datasets  (normalizaHon  or  standardizaHon)  will  
    completely  change  results.  While  this  itself  is  not  bad,  not  realizing  that  you  have  to  
    spend  extra  a4en(on  to  scaling  your  data  might  be  bad.    

    2016   Advantages  &  Disadvantages  of  k-­‐Means  and  Hierarchical  Clustering   3  


    Hierarchical  Clustering:  Advantages  and  Disadvantages  
    Advantages  
    •  Hierarchical  clustering  outputs  a  hierarchy,  ie  a  structure  that  is  more  informaHve  than  
    the  unstructured  set  of  flat  clusters  returned  by  k-­‐means.  Therefore,  it  is  easier  to  decide  
    on  the  number  of  clusters  by  looking  at  the  dendrogram  (see  suggesHon  on  how  to  cut  a  
    dendrogram  in  lab8).  
    •  Easy  to  implement  
    Disavantages  
    •  It  is  not  possible  to  undo  the  previous  step:  once  the  instances  have  been  assigned  to  a  
    cluster,  they  can  no  longer  be  moved  around.    
    •  Time  complexity:  not  suitable  for  large  datasets  
    •  IniHal  seeds  have  a  strong  impact  on  the  final  results  
    •  The  order  of  the  data  has  an  impact  on  the  final  results  
    •  Very  sensiHve  to  outliers  

    2016   Advantages  &  Disadvantages  of  k-­‐Means  and  Hierarchical  Clustering   4  


    The  end

    Advantages  &  Disadvantages  of  k-­‐Means  and  Hierarchical  


    2016   5  
    Clustering  

    You might also like