WUMprep4Weka Installation

From HypKNOWsys

WUMprep4Weka Navigation

README · Installation · Changelog


Contents

[edit]

Preliminary Remark

This documentation is not yet thoroughly tested and may change before the next official release. If you run into any problems or detect some mistake, please check first the online documentation on http://www.hypknowsys.org for updates and corrections!

[edit]

Prerequisites

[edit]

Perl

You need a Perl interpreter for running the WUMprep scripts. WUMprep has been developed in a Perl 5.8.x environment. Most Linux distributions install this by default, please check your distribution's documentation if necessary.

If you plan to use WUMprep and/or WUMprep4Weka on Windows, you probably have to get and install Perl on your own. Please refer to http://www.perl.org for links to Perl environments for both Windows and Linux/Unix and for detailed installation documentation.

[edit]

Perl Modules Required by WUMprep

Independent of the WUMprep4Weka release you whish to install, there is a number of Perl modules that are required by the WUMprep Perl scripts and that you must install before using WUMprep4Weka.

For instructions on installing Perl modules, please refer to any of the following HOWTOs:

If you are using ActivePerl on Windows, you might use PPM for installing missing modules (note that the URL line below might be wrapped):

The following modules used by WUMprep are part of the standard Perl 5.8.x environment:

  • File::Copy
  • FindBin
  • POSIX
  • Socket

In addition, WUMprep requires the following modules that are not part of a default Perl installation:

  • Date::Format
  • HTML::Template
  • HTTP::Date
  • URI::Escape

WUMprep4Weka must be able to find the Perl executable ("perl" on Linux, "perl.exe" on windows). For this, it searches the directories given in the PATH environment variable. Please make sure that Perl can be found this way.


[edit]

Java

While Weka runs on a Java 1.4 or later, WUMprep4Weka requires Java 1.5 (alternatively called "J2SE 1.5.0") or later.

WUMprep4Weka WONT RUN ON JAVA VERSIONS BEFORE 1.5!

If you don't already have the appropriate Java runtime environment installed (or the corresponding JDK, respectively), you can get it together with installation instructions for your platform from http://java.sun.com.


[edit]

Weka

WUMprep4Weka is a plugin for the Weka data mining environment, as it can be downloaded from http://www.cs.waikato.ac.nz/~ml/weka/.

WUMprep4Weka HAS BEEN DEVELOPED AND TESTED FOR WEKA RELEASE 3.5.1! This was the development release at the time of this writing. Furthermore, WUMprep4Weka requires a handful of minor modifications to official Weka releases <= 3.5.1. Thus, if you want to use the WUMprep4Weka plugin together with a separate version <= 3.5.1 of Weka, you have to get the Weka sources, apply the patches that come with WUMprep4Weka (see the directory misc/modifiedWekaFiles of the source distribution), and build Weka on your own!!! As long as you are not a Java developer, this is not trivial - better opt for the second alternative described below.

The easiest way to get WUMprep4Weka is to download the special "Weka-WP4W" release, which brings you the right Weka version with WUMprep and WUMprep4Weka already included. You can get it from http://www.hypknowsys.org.

[edit]

Installation

WUMprep4Weka is available in two "flavours": As a plugin for an existing Weka installation (the do-it-yourself way), and as part of a special Weka distribution (the convenient way). In the following, we describe how to install both of these variants.

For the moment, we strongly recommend to use the Weka-WP4W distribution!!!


[edit]

WUMprep4Weka Plugin Installation

ATTENTION: WUMprep4Weka requires a few patches applied to the official Weka sources <= version 3.5.1. If you have no idea what this means and how to compile the patched Weka on your own, please proceed to 2.2. and use the pre-packaged WUMprep4Weka release instead!

Follow these steps to install WUMprep4Weka as a plugin for your existing Weka installation (see the "Requirements" section for the required Weka version):

  1. Check whether your computer fits the requirements explained above.
  2. Get the Weka 3.5.1 sources (maybe a newer version works as well)
  3. Download and install the latest WUMprep release (>= 0.11.0) from http://www.hypknowsys.org.
  4. Download the latest WUMprep4Weka release from http://www.hypknowsys.org. The plugin comes as an archive that extracts into a subdirectory named "WUMprep4Weka-[version]".
    1. In the misc/patch directory of the WUMprep4Weka distribution, you'll find the file(s) containing the patches to the official Weka sources in "unified diff" format. You should know how to apply them. If not, please use the Weka-WP4W distribution.
  5. Weka (version >= 3.4) uses a file called "GenericPropertiesCreator.props" for defining which algorithms and filters to load. You must edit this file in order to make the WUMprep4Weka plugin available in Weka:
    • To "weka.filters.Filter", add "org.hypknowsys.wumprep4weka.filters"
    • To "weka.core.converters.Loader", add "org.hypknowsys.wumprep4weka.core.converters"
    • To "weka.core.converters.Saver", add "org.hypknowsys.wumprep4weka.core.converters"
    The default GenericPropertiesCreator.props file is located in the weka/gui/ directory of the Weka source distribution. According to Weka's documentation, you should place your modified copy of this file in your home directory. However, this does not seem to work reliably. Please refer to the Weka documentation for more information about installing plugins via the GenericPropertiesCreator.props file.
  6. When running Weka, add wumprep4weka[-releaseNo].jar to the Java classpath. If you start Weka the usual way via the GUI chooser, start it with the following command line: 
    java -classpath [path-to-wumprep4weka.jar]:$CLASSPATH -jar weka.jar
    Please refer to the Weka documentation for more information about starting Weka.
  7. Proceed as described under "Usage" in the README file.

[edit]

Weka-WP4W Installation

This describes how to install and run the "Weka-WP4W" ("Weka-WUMprep4Weka") distribution of Weka, which already includes both the WUMprep4Weka plugin and the appropriate version of the WUMprep scripts.

  1. Check whether your computer fits the requirements explained above.
  2. Download the latest Weka-WP4W release from http://www.hypknowsys.org (you'll find it via the WUMprep4Weka pages).
  3. Install Weka-WP4W by extracting the archive into a directory of your choice.
  4. MAKE SURE THAT PERL IS ON YOUR "PATH"!
  5. Run Weka-WP4W by issuing 
    java -jar weka-wp4w-[releaseNo].jar
  6. Proceed as described under "Usage" in the README file.

Retrieved from "http://hypknowsys.sourceforge.net/wiki/WUMprep4Weka_Installation"