ts-weka-node

WORK IN PROGRESS

Uses Weka v3.9.3.

Setup

1. Download the Weka 3.9.3 JAR from https://mvnrepository.com/artifact/nz.ac.waikato.cms.weka/weka-dev/3.9.3 and place the Weka JAR file in your src folder, e.g. under ./src/bin/weka-3.9.3.jar

2. If you want to use certain classes for attribute selection, download the attributeSelectionSearchMethods JAR from https://sourceforge.net/projects/weka/files/weka-packages/attributeSelectionSearchMethods1.0.7.zip. If you already downloaded the package using the WEKA Package Manager, you can find the file in <Home>\packages\attributeSelectionSearchMethods. Place the attributeSelectionSearchMethods JAR file in your src folder, e.g. under ./src/bin/attributeSelectionSearchMethods-1.0.7.jar.

3. If you want to use EvolutionarySearch for attribute selection, download the JAR from https://github.com/sebastian-luna-valero/EvolutionarySearch/releases/download/1.0.2/EvolutionarySearch.zip and place it in your src folder, e.g. under ./src/bin/EvolutionarySearch-1.0.2.jar

4. To use this plugin, create a new WekaLibraryService instance:

   import * as path from 'path';
   import {WekaLibraryService} from 'ts-weka-node/lib/services/weka-library.service';
   ...
   
   private binPath: string = path.join(__dirname, '../src/bin/');
   
   private wekaLibraryService: WekaLibraryService = new WekaLibraryService('./input', './output', this.binPath);
   

   The path for the attributeSelectionSearchMethods JAR is optional.

   The WekaLibraryService distinguishes between balanced and unbalanced datasets. It expects unbalanced and balanced datasets (as .arff) in the input folder:

   • ./input/datasets/balanced/...
   • ./input/datasets/unbalanced/...

   Additionally, if feature selection is performed, the initial files are placed in the following directory:

   • ./input/datasets/initial/...

   Whenever a file name is required as a parameter, it can be provided with or without the suffix .arff