Download presentation
Presentation is loading. Please wait.
1
Machine Learning with WEKA
2
Copyright: Martin Kramer (mkramer@wxs.nl)
WEKA: the bird Copyright: Martin Kramer 11/18/2018 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 11/18/2018 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) 11/18/2018 University of Waikato
5
WEKA only deals with “flat” files
@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 ... Flat file in ARFF format 11/18/2018 University of Waikato
6
WEKA only deals with “flat” files
@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 ... numeric attribute nominal attribute 11/18/2018 University of Waikato
7
11/18/2018 University of Waikato
8
11/18/2018 University of Waikato
9
11/18/2018 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, … 11/18/2018 University of Waikato
11
11/18/2018 University of Waikato
12
11/18/2018 University of Waikato
13
11/18/2018 University of Waikato
14
11/18/2018 University of Waikato
15
11/18/2018 University of Waikato
16
11/18/2018 University of Waikato
17
11/18/2018 University of Waikato
18
11/18/2018 University of Waikato
19
11/18/2018 University of Waikato
20
11/18/2018 University of Waikato
21
11/18/2018 University of Waikato
22
11/18/2018 University of Waikato
23
11/18/2018 University of Waikato
24
11/18/2018 University of Waikato
25
11/18/2018 University of Waikato
26
11/18/2018 University of Waikato
27
11/18/2018 University of Waikato
28
11/18/2018 University of Waikato
29
11/18/2018 University of Waikato
30
11/18/2018 University of Waikato
31
11/18/2018 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, … 11/18/2018 University of Waikato
33
11/18/2018 University of Waikato
34
11/18/2018 University of Waikato
35
11/18/2018 University of Waikato
36
11/18/2018 University of Waikato
37
11/18/2018 University of Waikato
38
11/18/2018 University of Waikato
39
11/18/2018 University of Waikato
40
11/18/2018 University of Waikato
41
11/18/2018 University of Waikato
42
11/18/2018 University of Waikato
43
11/18/2018 University of Waikato
44
11/18/2018 University of Waikato
45
11/18/2018 University of Waikato
46
11/18/2018 University of Waikato
47
11/18/2018 University of Waikato
48
11/18/2018 University of Waikato
49
11/18/2018 University of Waikato
50
11/18/2018 University of Waikato
51
11/18/2018 University of Waikato
52
11/18/2018 University of Waikato
53
11/18/2018 University of Waikato
54
11/18/2018 University of Waikato
55
11/18/2018 University of Waikato
56
11/18/2018 University of Waikato
57
11/18/2018 University of Waikato
58
11/18/2018 University of Waikato
59
11/18/2018 University of Waikato
60
11/18/2018 University of Waikato
61
11/18/2018 University of Waikato
62
11/18/2018 University of Waikato
63
11/18/2018 University of Waikato
64
11/18/2018 University of Waikato
65
11/18/2018 University of Waikato
66
11/18/2018 University of Waikato
67
11/18/2018 University of Waikato
68
11/18/2018 University of Waikato
69
11/18/2018 University of Waikato
70
11/18/2018 University of Waikato
71
11/18/2018 University of Waikato
72
11/18/2018 University of Waikato
73
11/18/2018 University of Waikato
74
11/18/2018 University of Waikato
75
11/18/2018 University of Waikato
76
11/18/2018 University of Waikato
77
11/18/2018 University of Waikato
78
11/18/2018 University of Waikato
79
11/18/2018 University of Waikato
80
11/18/2018 University of Waikato
81
11/18/2018 University of Waikato
82
11/18/2018 University of Waikato
83
11/18/2018 University of Waikato
84
11/18/2018 University of Waikato
85
11/18/2018 University of Waikato
86
11/18/2018 University of Waikato
87
11/18/2018 University of Waikato
88
11/18/2018 University of Waikato
89
11/18/2018 University of Waikato
90
11/18/2018 University of Waikato
91
11/18/2018 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 11/18/2018 University of Waikato
93
11/18/2018 University of Waikato
94
11/18/2018 University of Waikato
95
11/18/2018 University of Waikato
96
11/18/2018 University of Waikato
97
11/18/2018 University of Waikato
98
11/18/2018 University of Waikato
99
11/18/2018 University of Waikato
100
11/18/2018 University of Waikato
101
11/18/2018 University of Waikato
102
11/18/2018 University of Waikato
103
11/18/2018 University of Waikato
104
11/18/2018 University of Waikato
105
11/18/2018 University of Waikato
106
11/18/2018 University of Waikato
107
11/18/2018 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 11/18/2018 University of Waikato
109
11/18/2018 University of Waikato
110
11/18/2018 University of Waikato
111
11/18/2018 University of Waikato
112
11/18/2018 University of Waikato
113
11/18/2018 University of Waikato
114
11/18/2018 University of Waikato
115
11/18/2018 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 11/18/2018 University of Waikato
117
11/18/2018 University of Waikato
118
11/18/2018 University of Waikato
119
11/18/2018 University of Waikato
120
11/18/2018 University of Waikato
121
11/18/2018 University of Waikato
122
11/18/2018 University of Waikato
123
11/18/2018 University of Waikato
124
11/18/2018 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 11/18/2018 University of Waikato
126
11/18/2018 University of Waikato
127
11/18/2018 University of Waikato
128
11/18/2018 University of Waikato
129
11/18/2018 University of Waikato
130
11/18/2018 University of Waikato
131
11/18/2018 University of Waikato
132
11/18/2018 University of Waikato
133
11/18/2018 University of Waikato
134
11/18/2018 University of Waikato
135
11/18/2018 University of Waikato
136
11/18/2018 University of Waikato
137
11/18/2018 University of Waikato
138
Performing experiments
Experimenter makes it easy to compare the performance of different learning schemes For classification and regression problems Results can be written into file or database Evaluation options: cross-validation, learning curve, hold-out Can also iterate over different parameter settings Significance-testing built in! 11/18/2018 University of Waikato
139
11/18/2018 University of Waikato
140
11/18/2018 University of Waikato
141
11/18/2018 University of Waikato
142
11/18/2018 University of Waikato
143
11/18/2018 University of Waikato
144
11/18/2018 University of Waikato
145
11/18/2018 University of Waikato
146
11/18/2018 University of Waikato
147
11/18/2018 University of Waikato
148
11/18/2018 University of Waikato
149
11/18/2018 University of Waikato
150
11/18/2018 University of Waikato
151
11/18/2018 University of Waikato
152
The Knowledge Flow GUI New graphical user interface for WEKA
Java-Beans-based interface for setting up and running machine learning experiments Data sources, classifiers, etc. are beans and can be connected graphically Data “flows” through components: e.g., “data source” -> “filter” -> “classifier” -> “evaluator” Layouts can be saved and loaded again later 11/18/2018 University of Waikato
153
11/18/2018 University of Waikato
154
11/18/2018 University of Waikato
155
11/18/2018 University of Waikato
156
11/18/2018 University of Waikato
157
11/18/2018 University of Waikato
158
11/18/2018 University of Waikato
159
11/18/2018 University of Waikato
160
11/18/2018 University of Waikato
161
11/18/2018 University of Waikato
162
Can continue this... 11/18/2018 University of Waikato
163
11/18/2018 University of Waikato
164
11/18/2018 University of Waikato
165
11/18/2018 University of Waikato
166
11/18/2018 University of Waikato
167
11/18/2018 University of Waikato
168
11/18/2018 University of Waikato
169
11/18/2018 University of Waikato
170
11/18/2018 University of Waikato
171
11/18/2018 University of Waikato
172
11/18/2018 University of Waikato
173
Conclusion: try it yourself!
WEKA is available at 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 11/18/2018 University of Waikato
Similar presentations
