WEKA ： A Practical Machine Learning Tool

WEKA：A Practical Machine Learning Tool

Contents • 1.Introduction to Weka • 2.Explorer • 3.Other three main tools • 4.Conclusions • 5.Reference

Introduction – What is Weka? • Usage: The algorithms can either be applied: (1) directly to a dataset (without writing any codes); (2) called from your own Java code. • In nature: A flightless bird with an inquisitive nature found only on the islands of New Zealand. • Actually: A practical machine learning tool developed by the University of Waikato in New Zealand. It is short for Waikato Environment for Knowledge Analysis. • Definition: A collection of machine learning algorithms for data mining tasks. • Language: It is written in Java and runs on almost any platform.

Introduction – Weka consists of • Explorer • Experimenter • Knowledge flow • Simple Command Line Interface(CLI) • Other tools and Visualization • Java interface

Explorer • WEKA’s main graphical user interface • Gives access to all its facilities using menu selection and form filling.(Data-Preprocess/Classify/Cluster/Associate/Select Attributes/Visualize) • 1.Data • 2. Operations of Explorer with a Classification example.

Explorer – Data(1) Attribute-Class Attribute Instance Instances *.xls *.csv Tips：weather.arff ( C:/Program Files/Weka/data/ ) • From files: CSV, ARFF, C4.5…（no *.xls） • Data loaded from URL or DB

Explorer – Data(2) • ARFF(Attribute-Relation File Format) • @relation <relation-name> • @attribute <attribute-name> <datatype> • numeric (real or integer numbers) • <nominal-specification> • string • date [<date-format>] • @data • % notes • More details: • http://www.cs.waikato.ac.nz/ • ~ml/weka/arff.html

Explorer – Operations with an example Input data Data preprocess Choose classifier  Test options Run Result analysis

Explorer Summary Statistics Input data Select an attribute Visualization

Explorer Weka Filter Tune Parameters Apply the Filter Select a Filter

Explorer Tune Parameters Results Select a Classifier Decide how to evaluate Model list

Right-click on model to get Menu (save, visualize, etc)

Others – Experimenter Click it for Experimenter • Comparing different learning algorithms • ------on different datasets • ------with various parameter settings • ------and analyzing the performance statistics

Others – KnowledgeFlow Click it for KnowledgyFlow • The KnowledgeFlow provides an alternative to the Explorer as a graphical front end to Weka's core algorithms. • The KnowledgeFlow is a work in progress so some of the functionality from the Explorer is not yet available.

Others – Simple command line interface Click it for Simple CLI • All implementations of the algorithms have a uniform command-line interface. • java weka.classifiers.trees.J48 -t weather.arff

Conclusions 1.Explorer: Input data Data preprocess Choose classifier  Test options Run Result analysis 2.Experimenter: It is necessary for further studies. 3.Make full use of： • 1. Java tips; • 2. WekaManual.pdf; (C:/Program Files/Weka/ ) • 3. Play it yourself!

Reference • Mitchell, T. Machine Learning, 1997 McGraw Hill. • Ian H. Witten, Eibe Frank, Len Trigg, Mark Hall, Geoffrey Holmes, and Sally Jo Cunningham (1999). Weka: Practical machine learning tools and techniques with Java implementations. • Ian H. Witten, Eibe Frank (2005). Data Mining: Practical Machine Learning Tools and Techniques (Second Edition, 2005). San Francisco: Morgan Kaufmann • Weka Homepage: http://www.cs.waikato.ac.nz/~ml/weka/ • Wekawiki: http://weka.wikispaces.com/ • Weka on SourceForge.net: http://sourceforge.net/projects/weka • WekaManual.pdf (C:\Program Files\Weka-3-6\WekaManual.pdf)

WEKA ： A Practical Machine Learning Tool