Patrons of service
SOFTWARE
 The Rough Set Exploration System (RSES) is a freely available software system toolset for data exploration, classification support and knowledge discovery. Many of the RSES methods have originated from rough set theory introduced by Zdzis³aw Pawlak during the early 1980s. The Rough Set Exploration System (RSES) is a software system toolset that makes it possible to analyze tabular datasets utilizing various methods. In particular, many RSES methods are based on rough set theory. The first version of RSES and its companion RSESlib became available over a decade ago. After a number of modifications, improvements, and removal of detected bugs, RSES has been used in many applications. This version was prepared by the research team supervised by Professor Andrzej Skowron. Currently, the RSES R&D team consists of: Jan Bazan, Rafa³ Latkowski, Micha³ Miko³ajczyk, Nguyen Hung Son, Nguyen Sinh Hoa, Dominik ¦lêzak, Piotr Synak, Marcin Szczuka, Arkadiusz Wojna, Marcin Wojnarski, and Jakub Wróblewski. At the moment of this writing RSES version 2.2 is the most current. The RSES ver. 2.2 software and its computational kernel - the RSESlib 3.0 library - maintains all advantages of previous versions. The algorithms have been redesigned to provide better flexibility, extended functionality and the ability to process massive data sets. New algorithms added to the library reflect the current state of our research in classification methods originating in rough sets theory. Improved construction of library allows further extensions and supports augmentation of RSESlib into different data analysis tools. Today RSES is freely distributed for non-commercial purposes. Anybody can download it from the Web site http://logic.mimuw.edu.pl/~rses. RSES is a computer software system developed for the purpose of analyzing data. The data is assumed to be in the form of information system or decision table. The main step in the process of data analysis with RSES is the construction and evaluation of classifiers. Classification algorithms, or classifiers, are algorithms that permit us to repeatedly make a forecast in new situation on the basis of accumulated knowledge. In our case the knowledge is embedded in the structure of classifier which itself is constructed (learned) from data. RSES utilizes classification algorithms using elements of rough set theory, instance based learning, artificial neural networks and others. The construction of classifier is usually preceded by several initial steps. First, the data for analysis has to be loaded/imported into RSES. RSES can accept several input formats. Once the data is loaded, the user can examine it using provided visualization and statistics tools. In order to have a better chance for constructing (learning) a proper classifier, it is frequently advisable to transform the initial data set. Such transformation, usually referred to as preprocessing may consist of several steps. RSES supports preprocessing methods which make it possible to manage missing parts in data, discretize numeric attributes, and create new attributes. Once the data is preprocessed we may be interested in learning about its internal structure. By using classical rough set concepts such as reducts, dynamic reducts and positive region one may pinpoint dependencies that occur in our data set. Knowledge of reducts may lead to reduction of data by removing some of the redundant attributes. Reducts can also provide essential hints for the parameter setting during classifier construction. To simplify the use of RSES algorithms and make it more intuitive the RSES graphical user interface was constructed. It is directed towards ease of use and visual representation of workflow. Project interface window consists of two parts. The visible part is the project workspace with icons representing objects created during blue computation. Behind the project window there is the history window, reachable via tab,and dedicated to messages, status reports, errors and warnings. An important, recently added GUI feature is the possibility to display some statistical information about tables, rules and reducts in a graphical form. Behind the front-end that is visible to the user, there is RSES computational kernel. This most essential part of the system is built around the library of methods known as RSESlib ver. 3.0. The library is mostly written in Java but, it also uses a part that was implemented using C++. The C++ part is the legacy of previous RSESlib versions and contains those algorithms that could only lose optimality if re-implemented in Java. Currently, it is possible to install RSES in Microsoft Windows 95/98/2000/XP and in Linux/i386. The computer on which the RSES is installed has to be equipped with Java Runtime Environment. During operation certain functions in RSES may read and write information to/from files. Most of the files that can be read or written are regular ASCII text files. A particular sub-types can be distinguished by reviewing the contents or identifying file extensions (if used). As the whole system is about analyzing tabular data, it is equipped with abilities to read several tabular data formats. At the time of this writing the system can import text files formatted for old version of RSES (RSES1 format), Rosetta, and Weka systems. Naturally, there exists native RSES2 file format used to store data tables. The RSES user can save and retrieve data entities created during experiment, such as rule sets, reduct sets etc. The option of saving the whole workspace (project) in a single file is also provided. The project layout together with underlying data structures is stored using dedicated, optimized binary file format. We would like to express our gratitude to all the current and previous members and supporters of RSES development team, in particular the creators of Rosetta - Aleksander Ohrn and Jan Komorowski. Over the years development of our software was significantly supported by several national and international research grants. Currently RSES is supported by the grant 3T11C00226 from Polish Ministry of Scientific Research and Information Technology. |