# Rseslib & QMAK
Tools, algorithms and data structures in Java for rough sets and explainable machine learning.

The project includes algorithms for lower and upper approximation, discernibility matrix, reducts, decision rules and wide range of discretization and classification algorithms, including KNN classifier optimizing automatically the distance measure and the number of voting neighbors, a neural network with automated selection of network structure, and dedicated to imbalanced data RIONIDA with multidimensional self-optimization. The project includes also QMAK - a tool for interacting with machine learning models and visualizing classification process, and Simple Grid Manager for running experiments on many computers or cores. For more information visit https://rseslib.mimuw.edu.pl.

The project is built with Java Development Kit 8 and Maven. The following command will build *rseslib-\<version\>.jar* and copy other needed libraries to Maven target directory:
```
mvn package
```

Rseslib can be used in the following ways:

### 1. QMAK
QMAK is a GUI tool included in Rseslib. 5-minute video demonstrating the tool is available at [QMAK website](http://rseslib.mimuw.edu.pl/qmak). The simplest way to run QMAK is to download the package from the website, unpack it and run *qmak.sh* (on Linux) or *qmak.bat* (on Windows). See the chapter *QMAK: Interaction wit classifers and their visualization* in [Rseslib User Guide](https://rseslib.mimuw.edu.pl/rseslib.pdf) and *Help* in the main menu of the appliction for information how to use the tool. If you run QMAK from the source add jcommon-0.9.6.jar, jfreechart-0.9.21.jar and weka jar version 3.8.x to the classpath, for example:
```
java -cp jcommon-0.9.6.jar:jfreechart-0.9.21.jar:weka.jar:rseslib.jar rseslib.qmak.QmakMain
```

### 2. Java library
[Rseslib 3 algorithms](https://rseslib.mimuw.edu.pl/algorithms.html) provide a brief list of algorithms available in the library and [Rseslib User Guide](https://rseslib.mimuw.edu.pl/rseslib.pdf) is the main source of information how to use Rseslib components within Java code.

### 3. Command line
Rseslib includes command-line programs evaluating attributes, computing reducts or rules or running experiments with Rseslib classifiers. See the chapter *Command line programs* in [Rseslib User Guide](https://rseslib.mimuw.edu.pl/rseslib.pdf) for information how to run them. While starting a program, add weka jar version 3.8.x to classpath to make it work on ARFF data files, for example:
```
java -cp rseslib.jar:weka.jar rseslib.example.ComputeReducts data/iris.arff iris.reducts
```

### 4. Simple Grid Manager
Simple Grid Manager is a tool included in Rseslib for running experiments with Rseslib classifiers on many computers or cores. See the chapter *SGM: Computing many experiments on many computers/cores* in [Rseslib User Guide](https://rseslib.mimuw.edu.pl/rseslib.pdf) for information how to configure experiments and run the tool.

### 5. WEKA platform
4 selected classifiers (Rough Set based, K Nearest Neighbors, K Nearest Neighbors with Local Metric Induction, and RIONIDA) are available in WEKA. See the chapter *WEKA* in [Rseslib User Guide](https://rseslib.mimuw.edu.pl/rseslib.pdf) for information how to install the Rseslib package using WEKA package manager and where to find Rseslib classifiers in WEKA catalog.