Software

Three machine learning toolkits are shortly described and available below: CORElearn, SemiArtificial and ExplainPrediction. The software is available under the GNU public license (GPLv3 or later). Other licensing options are available.

CORElearn

The system supports feature evaluation, decision, regression trees, random forests, OrdEval algorithm, explanation of predictions and some machine learning utilities. The source code and binaries of the learning system CORElearn can be downloaded as:

port to R system available via CRAN: sources and also binaries for different platforms (regularly tested tested for Windows, Linux, and OS X)
the latest development version updated 08 February, 2020: CORElearn_1.56.0.tar.gz and CORElearn_1.56.0.zip (Windows binary)
standalone version (version 0.9.44, 25 January, 2015) consists of the same C++ code, plus compilation instructions, parameters file, data format description, and same sample testing data. It is tested to compile with gcc and msvc and contains Windows executable. It may be older than the R port, mail me if you need a fresh one. To compile the sources read the ReadMe.htm file first. Supported data input formats are C4.5, C5, and native format. The system is tested to compile and run on Windows and Linux. The documentation is available through R, just install the package and type ?CORElearn at R prompt.

Some of the functionality and features are described in my papers. The R package contains the documentation which applies also to the standalone version.

SemiArtificial

The system for generating new data based on the properties of an existing data set is available via CRAN. Current version is semiArtificial 2.4.1 (23 September 2021). The working principles are described in my papers.

ExplainPrediction

The package allows explanation of predictions of a given prediction model. Two explanation methods (EXPLAIN and IME) and two levels of prediction explanations are available:

an instance level explanation, which shows contributions of features to the decision of the model for individual instance,
a model level explanation, which average explanations of individual instances to create a summary explanation of the whole model.

The package is available via CRAN. Current development version is ExplainPrediction 1.3.0 (07 January, 2018). The working principles are described in my papers.

Home page of Marko Robnik Šikonja.