A Taxonomy and Experimental Study on Prototype Generation for Nearest Neighbor Classification - Complementary Material

This Website contains additional material to the SCI2S research paper on Prototype Generation

I. Triguero, J. Derrac, S. García and F.Herrera, A Taxonomy and Experimental Study on Prototype Generation for Nearest Neighbor Classification . IEEE Transactions on Systems, Man, and Cybernetics--Part C: Applications and Reviews 42 (1) (2012) 86-100, doi: 10.1109/TSMCC.2010.2103939 PDF Icon

Summary:

  1. Abstract
  2. Experimental study
  3. JAVA code for PG methods
    1. Source Code
    2. Data sets
    3. KEEL version with PG package

JAVA code for PG methods

In this section, we provide the full package of algorithms employed in the experimental study, available for public use. It is composed by the source code of the methods, the data sets, and a modified version of KEEL with the PG package integrated.

Source code

The source code of the PG package can be downloaded from here. (FALTA CODIGO)

It is written in the Java programming language. Although it is developed to be run inside the KEEL environment, it can be also executed on a standard Java Virtual Machine. To do this, it is only needed to place the training datasets at the same folder, and write a script which contains al the relevant parameters of the algorithms (see the KEEL reference manual, section 3; located under the KEEL Software Tool item of the main menu).

Data sets

The data sets employed in the experimental study can be downloaded from here.

They are represented in KEEL data format, which is fairly similar to ARFF (WEKA data format). For each data set, we provide a file containing the full dataset, and 10 additional files corresponding to the data set partitioned (ready to be employed in a 10-fold cross validation procedure)

KEEL version with PG package

The KEEL version with PG package can be downloaded from here.

The complete KEEL user manual can be downloaded from here.

Essentially, to generate a experiment with this modified version of KEEL, it is needed to perform the following steps:

- Execute the KEEL GUI with the command: java -jar GraphInterKeel.jar.

- Select Experiments.

- Select Classification (it is recommended the use of 10-fold cross validation, as set by default).

- Select the data sets desired to use, and click on the main window to place them.

- Click on the second icon of the left bar, select the IG method desired, and click on the main window to place it.

- Click on the last icon of the left bar, and create some arcs which joins all the nodes of the experiments.

- Click on the Run Experiment icon, located at the top bar.

This way, it is possible to generate an experiment ready to be executed on any machine with a Java Virtual Machine installed.