<< Homepage
ART Forests Software
artforests-0.1 is a standalone program (architecture) configured for windows.
Although the program can be used for training/predicting on standard multi-rank/label ranking datasets, version 0.1 was implemented for demonstrating the general idea of the ART forests algorithm.
Please let me know if you have any questions.
A more comprehensive package is available for download here
Installation
Unzip artforests-0.1.zip to C:\artforests-0.1
Configuration
Config training
Edit C:\artforests-0.1\config\trainConfig.xml
Config testing
Edit C:\artforests-0.1\config\testConfig.xml
How to Use artforests-0.1
Click train.bat to build an ARTF model
Click predict.bat to test a model and output predictions to a CSV file.
Programmatic use /Java API
Unzip artforests-0.1.zip
Please see BuildModel.java (screenshot) and Predict.java in 'src/artforests/app/' for an example.
External libraries
Unzip artforests-0.1.zip
Included in 'lib'.
Supplementary materials
Click here to download File 1, which includes:
Pages 2 - 3
Ranked lists of the 20 algorithms over 466 binary classification datasets;
Pearson's correlation scores between ranking positions of each pair of algorithms
Pages 4 - 5
Parameters of the 20 WEKA algorithms for PSO to optimize
Page 6
Properties of the 466 datasets used for experiments
Pages 7 - 8
The 80 base-level SIL meta-features
Misc.
Lonora Jessamine
Phil Cutler
Last updated: 07/2013