Page 1
Department of Computer Science,
University of Waikato, New Zealand
Eibe Frank
WEKA: A Machine
Learning Toolkit
The Explorer
• Classification and
Regression
• Clustering
• Association Rules
• Attribute Selection
• Data Visualization
The Experimenter
The Knowledge
Flow GUI
Conclusions
Machine Learning with WEKA
Page 2
2/22/2011 University of Waikato 2
WEKA: the bird
Copyright: Martin Kramer ([email protected] )
Page 3
2/22/2011 University of Waikato 3
WEKA: the software
Machine learning/data mining software written in
Java (distributed under the GNU Public License)
Used for research, education, and applications
Complements “Data Mining” by Witten & Frank
Main features:
Comprehensive set of data pre-processing tools,
learning algorithms and evaluation methods
Graphical user interfaces (incl. data visualization)
Environment for comparing learning algorithms
Page 4
2/22/2011 University of Waikato 4
WEKA: versions
There are several versions of WEKA:
WEKA 3.0: “book version” compatible with
description in data mining book
WEKA 3.2: “GUI version” adds graphical user
interfaces (book version is command-line only)
WEKA 3.3: “development version” with lots of
improvements
This talk is based on the latest snapshot of WEKA
3.3 (soon to be WEKA 3.4)
Page 5
2/22/2011 University of Waikato 5
@relation heart-disease-simplified
@attribute age numeric
@attribute sex { female, male}
@attribute chest_pain_type { typ_angina, asympt, non_anginal, atyp_angina}
@attribute cholesterol numeric
@attribute exercise_induced_angina { no, yes}
@attribute class { present, not_present}
@data
63,male,typ_angina,233,no,not_present
67,male,asympt,286,yes,present
67,male,asympt,229,yes,present
38,female,non_anginal,?,no,not_present
...
WEKA only deals with “flat” files
Page 6
2/22/2011 University of Waikato 6
@relation heart-disease-simplified
@attribute age numeric
@attribute sex { female, male}
@attribute chest_pain_type { typ_angina, asympt, non_anginal, atyp_angina}
@attribute cholesterol numeric
@attribute exercise_induced_angina { no, yes}
@attribute class { present, not_present}
@data
63,male,typ_angina,233,no,not_present
67,male,asympt,286,yes,present
67,male,asympt,229,yes,present
38,female,non_anginal,?,no,not_present
...
WEKA only deals with “flat” files
Page 7
2/22/2011 University of Waikato 7
Page 8
2/22/2011 University of Waikato 8
Page 9
2/22/2011 University of Waikato 9
Page 10
2/22/2011 University of Waikato 10
Explorer: pre-processing the data
Data can be imported from a file in various
formats: ARFF, CSV, C4.5, binary
Data can also be read from a URL or from an SQL
database (using JDBC)
Pre-processing tools in WEKA are called “filters”
WEKA contains filters for:
Discretization, normalization, resampling, attribute
selection, transforming and combining attributes, …
Page 11
2/22/2011 University of Waikato 11
Page 12
2/22/2011 University of Waikato 12
Page 13
2/22/2011 University of Waikato 13
Page 14
2/22/2011 University of Waikato 14
Page 15
2/22/2011 University of Waikato 15
Page 16
2/22/2011 University of Waikato 16
Page 17
2/22/2011 University of Waikato 17
Page 18
2/22/2011 University of Waikato 18
Page 19
2/22/2011 University of Waikato 19
Page 20
2/22/2011 University of Waikato 20
Page 21
2/22/2011 University of Waikato 21
Page 22
2/22/2011 University of Waikato 22
Page 23
2/22/2011 University of Waikato 23
Page 24
2/22/2011 University of Waikato 24
Page 25
2/22/2011 University of Waikato 25
Page 26
2/22/2011 University of Waikato 26
Page 27
2/22/2011 University of Waikato 27
Page 28
2/22/2011 University of Waikato 28
Page 29
2/22/2011 University of Waikato 29
Page 30
2/22/2011 University of Waikato 30
Page 31
2/22/2011 University of Waikato 31
Page 32
2/22/2011 University of Waikato 32
Explorer: building “classifiers”
Classifiers in WEKA are models for predicting
nominal or numeric quantities
Implemented learning schemes include:
Decision trees and lists, instance-based classifiers,
support vector machines, multi-layer perceptrons,
logistic regression, Bayes’ nets, …
“Meta”-classifiers include:
Bagging, boosting, stacking, error-correcting output
codes, locally weighted learning, …
Page 33
2/22/2011 University of Waikato 33
Page 34
2/22/2011 University of Waikato 34
Page 35
2/22/2011 University of Waikato 35
Page 36
2/22/2011 University of Waikato 36
Page 37
2/22/2011 University of Waikato 37
Page 38
2/22/2011 University of Waikato 38
Page 39
2/22/2011 University of Waikato 39
Page 40
2/22/2011 University of Waikato 40
Page 41
2/22/2011 University of Waikato 41
Page 42
2/22/2011 University of Waikato 42
Page 43
2/22/2011 University of Waikato 43
Page 44
2/22/2011 University of Waikato 44
Page 45
2/22/2011 University of Waikato 45
Page 46
2/22/2011 University of Waikato 46
Page 47
2/22/2011 University of Waikato 47
Page 48
2/22/2011 University of Waikato 48
Page 49
2/22/2011 University of Waikato 49
Page 50
2/22/2011 University of Waikato 50
Page 51
2/22/2011 University of Waikato 51
Page 52
2/22/2011 University of Waikato 52
Page 53
2/22/2011 University of Waikato 53
Page 54
2/22/2011 University of Waikato 54
Page 55
2/22/2011 University of Waikato 55
Page 56
2/22/2011 University of Waikato 56
Page 57
2/22/2011 University of Waikato 57
Page 58
2/22/2011 University of Waikato 58
Page 59
2/22/2011 University of Waikato 59
Page 60
2/22/2011 University of Waikato 60
Page 61
2/22/2011 University of Waikato 61
Page 62
2/22/2011 University of Waikato 62
Page 63
2/22/2011 University of Waikato 63
Page 64
2/22/2011 University of Waikato 64
Page 65
2/22/2011 University of Waikato 65QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture.
Page 66
2/22/2011 University of Waikato 66QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture.
Page 67
2/22/2011 University of Waikato 67QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture.
Page 68
2/22/2011 University of Waikato 68
Page 69
2/22/2011 University of Waikato 69
Page 70
2/22/2011 University of Waikato 70
Page 71
2/22/2011 University of Waikato 71
Page 72
2/22/2011 University of Waikato 72
Page 73
2/22/2011 University of Waikato 73
Page 74
2/22/2011 University of Waikato 74
Page 75
2/22/2011 University of Waikato 75
Quic k Time™ and a TIFF (LZW) dec ompres s or are needed to s ee this pic ture.
Page 76
2/22/2011 University of Waikato 76
Page 77
2/22/2011 University of Waikato 77
Page 78
2/22/2011 University of Waikato 78
Page 79
2/22/2011 University of Waikato 79
Page 80
2/22/2011 University of Waikato 80
QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture.
Page 81
2/22/2011 University of Waikato 81
QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture.
Page 82
2/22/2011 University of Waikato 82
Page 83
2/22/2011 University of Waikato 83
QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture.
Page 84
2/22/2011 University of Waikato 84
Page 85
2/22/2011 University of Waikato 85
Page 86
2/22/2011 University of Waikato 86
Page 87
2/22/2011 University of Waikato 87
Page 88
2/22/2011 University of Waikato 88
Page 89
2/22/2011 University of Waikato 89
Page 90
2/22/2011 University of Waikato 90
Page 91
2/22/2011 University of Waikato 91
Page 92
2/22/2011 University of Waikato 92
Explorer: clustering data
WEKA contains “clusterers” for finding groups of
similar instances in a dataset
Implemented schemes are:
k-Means, EM, Cobweb, X-means, FarthestFirst
Clusters can be visualized and compared to “true”
clusters (if given)
Evaluation based on loglikelihood if clustering
scheme produces a probability distribution
Page 93
2/22/2011 University of Waikato 93
Page 94
2/22/2011 University of Waikato 94
Page 95
2/22/2011 University of Waikato 95
Page 96
2/22/2011 University of Waikato 96
Page 97
2/22/2011 University of Waikato 97
Page 98
2/22/2011 University of Waikato 98
Page 99
2/22/2011 University of Waikato 99
Page 100
2/22/2011 University of Waikato 100
Page 101
2/22/2011 University of Waikato 101
Page 102
2/22/2011 University of Waikato 102
Page 103
2/22/2011 University of Waikato 103
Page 104
2/22/2011 University of Waikato 104
Page 105
2/22/2011 University of Waikato 105
Page 106
2/22/2011 University of Waikato 106
Page 107
2/22/2011 University of Waikato 107
Page 108
2/22/2011 University of Waikato 108
Explorer: finding associations
WEKA contains an implementation of the Apriori
algorithm for learning association rules
Works only with discrete data
Can identify statistical dependencies between
groups of attributes:
milk, butter bread, eggs (with confidence 0.9 and
support 2000)
Apriori can compute all rules that have a given
minimum support and exceed a given confidence
Page 109
2/22/2011 University of Waikato 109
Page 110
2/22/2011 University of Waikato 110
Page 111
2/22/2011 University of Waikato 111
Page 112
2/22/2011 University of Waikato 112
Page 113
2/22/2011 University of Waikato 113
Page 114
2/22/2011 University of Waikato 114
Page 115
2/22/2011 University of Waikato 115
Page 116
2/22/2011 University of Waikato 116
Explorer: attribute selection
Panel that can be used to investigate which
(subsets of) attributes are the most predictive ones
Attribute selection methods contain two parts:
A search method: best-first, forward selection,
random, exhaustive, genetic algorithm, ranking
An evaluation method: correlation-based, wrapper,
information gain, chi-squared, …
Very flexible: WEKA allows (almost) arbitrary
combinations of these two
Page 117
2/22/2011 University of Waikato 117
Page 118
2/22/2011 University of Waikato 118
Page 119
2/22/2011 University of Waikato 119
Page 120
2/22/2011 University of Waikato 120
Page 121
2/22/2011 University of Waikato 121
Page 122
2/22/2011 University of Waikato 122
Page 123
2/22/2011 University of Waikato 123
Page 124
2/22/2011 University of Waikato 124
Page 125
2/22/2011 University of Waikato 125
Explorer: data visualization
Visualization very useful in practice: e.g. helps to
determine difficulty of the learning problem
WEKA can visualize single attributes (1-d) and
pairs of attributes (2-d)
To do: rotating 3-d visualizations (Xgobi-style)
Color-coded class values
“Jitter” option to deal with nominal attributes (and
to detect “hidden” data points)
“Zoom-in” function
Page 126
2/22/2011 University of Waikato 126
Page 127
2/22/2011 University of Waikato 127
Page 128
2/22/2011 University of Waikato 128
Page 129
2/22/2011 University of Waikato 129
Page 130
2/22/2011 University of Waikato 130
Page 131
2/22/2011 University of Waikato 131
Page 132
2/22/2011 University of Waikato 132
Page 133
2/22/2011 University of Waikato 133
Page 134
2/22/2011 University of Waikato 134
Page 135
2/22/2011 University of Waikato 135
Page 136
2/22/2011 University of Waikato 136
Page 137
2/22/2011 University of Waikato 137
Page 138
2/22/2011 University of Waikato 138
Conclusion: try it yourself!
WEKA is available at
http://www.cs.waikato.ac.nz/ml/weka
Also has a list of projects based on WEKA
WEKA contributors:
Abdelaziz Mahoui, Alexander K. Seewald, Ashraf M. Kibriya, Bernhard
Pfahringer , Brent Martin, Peter Flach, Eibe Frank ,Gabi Schmidberger
,Ian H. Witten , J. Lindgren, Janice Boughton, Jason Wells, Len Trigg,
Lucio de Souza Coelho, Malcolm Ware, Mark Hall ,Remco Bouckaert ,
Richard Kirkby, Shane Butler, Shane Legg, Stuart Inglis, Sylvain Roy,
Tony Voyle, Xin Xu, Yong Wang, Zhihai Wang