an introduction to weka explorer -...

In part from: Yizhou Sun 2008

An Introduction to WEKA Explorer

What is WEKA? � Waikato Environment for Knowledge Analysis

�  It’s a data mining/machine learning tool developed by Department of Computer Science, University of Waikato, New Zealand.

� Weka is also a bird found only on the islands of New Zealand.

2 22/01/16

Download and Install WEKA �  Website:

http://www.cs.waikato.ac.nz/~ml/weka/index.html �  Support multiple platforms (written in java):

� Windows, Mac OS X and Linux

3 22/01/16

Main Features �  49 data preprocessing tools �  76 classification/regression algorithms �  8 clustering algorithms �  3 algorithms for finding association rules �  15 attribute/subset evaluators + 10 search algorithms

for feature selection

4 22/01/16

Main GUI �  Three graphical user interfaces

�  “The Explorer” (exploratory data analysis) �  “The Experimenter” (experimental

environment) �  “The KnowledgeFlow” (new process model

inspired interface) �  Simple CLI- provides users without a graphic

interface option the ability to execute commands from a terminal window

5 22/01/16

Explorer �  The Explorer:

�  Preprocess data � Classification � Clustering � Association Rules � Attribute Selection � Data Visualization

�  References and Resources

6 22/01/16

22/01/16 7

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, …

22/01/16 8

@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

22/01/16 9

@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

22/01/16 University of Waikato 10

22/01/16 University of Waikato 11

IRIS dataset �  5 attributes, one is the classification �  3 classes: setosa, versicolor, virginica

22/01/16 University of Waikato 15

Attribute data �  Min, max and average value of attributes �  distribution of values :number of items for which:

�  class: distribution of attribute values in the classes

ai = v j | ai ∈ A,v j ∈V

22/01/16 University of Waikato 18

22/01/16 University of Waikato 19

22/01/16 University of Waikato 20

22/01/16 University of Waikato 21

Filtering attributes �  Once the initial data has been selected and loaded the user

can select options for refining the experimental data. �  The options in the preprocess window include selection of

optional filters to apply and the user can select or remove different attributes of the data set as necessary to identify specific information.

�  The user can modify the attribute selection and change the relationship among the different attributes by deselecting different choices from the original data set.

�  There are many different filtering options available within the preprocessing window and the user can select the different options based on need and type of data present.

22/01/16 University of Waikato 23

22/01/16 University of Waikato 24

22/01/16 University of Waikato 25

22/01/16 University of Waikato 26

22/01/16 University of Waikato 27

22/01/16 University of Waikato 28

22/01/16 University of Waikato 29

22/01/16 University of Waikato 30

22/01/16 University of Waikato 31

Discretizes in 10 bins of equal frequency

22/01/16 University of Waikato 32

Discretizes in 10 bins of equal frequency

22/01/16 University of Waikato 33

Discretizes in 10 bins of equal frequency

22/01/16 University of Waikato 34

22/01/16 University of Waikato 35

22/01/16 University of Waikato 36

22/01/16 37

Explorer: building “classifiers” �  “Classifiers” in WEKA are machine learning algorithmsfor

predicting nominal or numeric quantities �  Implemented learning algorithms include:

� Conjunctive rules, decision trees and lists, instance-based classifiers, support vector machines, multi-layer perceptrons, logistic regression, Bayes’ nets, …

Explore Conjunctive Rules learner

Need a simple dataset with few attributes , let’s select the weather dataset

Select a Classifier

Select training method

Right-click to select parameters

numAntds= number of antecedents, -1= empty rule

Select numAntds=10

Results are shown in the right window (can be scrolled)

Can change the right hand side variable

Performance data

Decision Trees with WEKA

22/01/16 University of Waikato 47

22/01/16 University of Waikato 48

22/01/16 University of Waikato 49

22/01/16 University of Waikato 50

22/01/16 University of Waikato 51

22/01/16 University of Waikato 52

22/01/16 University of Waikato 53

22/01/16 University of Waikato 54

22/01/16 University of Waikato 55

22/01/16 University of Waikato 56

22/01/16 University of Waikato 57

22/01/16 University of Waikato 58

22/01/16 University of Waikato 59

22/01/16 University of Waikato 60

22/01/16 University of Waikato 61

22/01/16 University of Waikato 62

22/01/16 University of Waikato 63

22/01/16 University of Waikato 64

22/01/16 University of Waikato 65

22/01/16 University of Waikato 66

22/01/16 University of Waikato 67

22/01/16 University of Waikato 68

22/01/16 69

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

gennaio 22, 2016 70

�  Given k, the k-means algorithm is implemented in four steps:

�  Partition objects into k nonempty subsets

� Compute seed points as the centroids of the clusters of the

current partition (the centroid is the center, i.e., mean point, of the cluster)

� Assign each object to the cluster with the nearest seed point

� Go back to Step 2, stop when no more new assignment

The K-Means Clustering Method

right click: visualize cluster assignement

22/01/16 73

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

gennaio 22, 2016 74

Basic Concepts: Frequent Patterns

�  itemset: A set of one or more items �  k-itemset X = {x1, …, xk} �  (absolute) support, or, support count of X:

Frequency or occurrence of an itemset X

�  (relative) support, s, is the fraction of transactions that contains X (i.e., the probability that a transaction contains X)

�  An itemset X is frequent if X’s support is no less than a minsup threshold

Customer buys diaper

Customer buys both

Customer buys beer

Tid Items bought

10 Beer, Nuts, Diaper

20 Beer, Coffee, Diaper

30 Beer, Diaper, Eggs

40 Nuts, Eggs, Milk

50 Nuts, Coffee, Diaper, Eggs, Milk

gennaio 22, 2016 75

Basic Concepts: Association Rules

�  Find all the rules X à Y with minimum support and confidence �  support, s, probability that a

transaction contains X ∪ Y �  confidence, c, conditional probability

that a transaction having X also contains Y

Let minsup = 50%, minconf = 50%

Freq. Pat.: Beer:3, Nuts:3, Diaper:4, Eggs:3, {Beer, Diaper}:3

Customer buys diaper

Customer buys both

Customer buys beer

Nuts, Eggs, Milk 40 Nuts, Coffee, Diaper, Eggs, Milk 50

Beer, Diaper, Eggs 30

Beer, Coffee, Diaper 20

Beer, Nuts, Diaper 10

Items bought Tid

n  Association rules: (many more!) n  Beer à Diaper (60%, 100%) n  Diaper à Beer (60%, 75%)

22/01/16 University of Waikato 77

22/01/16 University of Waikato 78

22/01/16 University of Waikato 79

22/01/16 University of Waikato 80

22/01/16 University of Waikato 81

1.  adoption-of-the-budget-resolution=y physician-fee-freeze=n 219 ==> Class=democrat 219 conf:(1)

2.  adoption-of-the-budget-resolution=y physician-fee-freeze=n aid-to-nicaraguan-contras=y 198 ==> Class=democrat 198 conf:(1)

3.  physician-fee-freeze=n aid-to-nicaraguan-contras=y 211 ==> Class=democrat 210 conf:(1)

ecc.

22/01/16 83

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

22/01/16 University of Waikato 84

22/01/16 University of Waikato 85

22/01/16 University of Waikato 86

22/01/16 University of Waikato 87

22/01/16 University of Waikato 88

22/01/16 University of Waikato 89

22/01/16 University of Waikato 90

22/01/16 University of Waikato 91

22/01/16 92

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

22/01/16 University of Waikato 94

22/01/16 University of Waikato 95

22/01/16 University of Waikato 96

22/01/16 University of Waikato 97

22/01/16 University of Waikato 98

click on a cell

22/01/16 University of Waikato 99

22/01/16 University of Waikato 100

22/01/16 University of Waikato 101

22/01/16 University of Waikato 102

22/01/16 University of Waikato 103

References and Resources �  References:

�  WEKA website: http://www.cs.waikato.ac.nz/~ml/weka/index.html

�  WEKA Tutorial: �  Machine Learning with WEKA: A presentation demonstrating all graphical user

interfaces (GUI) in Weka. �  A presentation which explains how to use Weka for exploratory data mining.

�  WEKA Data Mining Book: �  Ian H. Witten and Eibe Frank, Data Mining: Practical Machine Learning Tools

and Techniques (Second Edition) �  WEKA Wiki: http://weka.sourceforge.net/wiki/index.php/

Main_Page �  Others:

�  Jiawei Han and Micheline Kamber, Data Mining: Concepts and Techniques, 2nd ed.