ROSECON is a software system supporting users with automated discovering net models from data tables as well as predicting model changes in time. It is run on PC computers under Windows operating system.
The acronym ROSECON comes from "ROugh SEts and CONcurrency". ROSECON offers a user-friendly GUI environment for triggering computations and displaying results. ROSECON is the successor of TAS and PN-tools systems created by Zbigniew Suraj's group in the 1990s. ROSECON is being developed in Chair of Computer Science Foundations at the University of Information Technology and Management in RzeszÃ³w, Poland, under the supervision of Zbigniew Suraj. The main performer directly involved in the process of designing and programming is Krzysztof Pancerz.
Information systems (in Pawlak's sense) and dynamic information systems (introduced by Z. Suraj) can be used to represent the knowledge about behaviour of concurrent systems. The idea of concurrent system representation by information systems is due to Z. Pawlak. Also, other researchers, especially A. Skowron, Z. Suraj, J.F. Peters, and K. Pancerz have contributed to the development of that idea. An information system represented by a data table includes the knowledge about global states of a given concurrent system. The columns of a table are labelled with names of processes of a given concurrent system, therefore attributes can be interpreted as processes. Each row labelled with an object includes a record of local states of processes. It can be interpreted as a global state of a given system. With each process, the set of its internal (local) states is associated. The description of concurrent systems by classic information systems does not take their behaviour into consideration, i.e., an information system includes only the knowledge about global states observed in a given concurrent system, but there is a lack of information about transitions between these states. Therefore, it can be assumed that all transitions between observed states are possible. However, a dynamic information system additionally includes information about transitions between global states observed in a given concurrent system. The dynamics is expressed by a transition relation defined in a dynamic information system and the term of a dynamic information system should be understood in this sense. Then, the behaviour of a concurrent system is presented in the form of two tables. The first table represents global states of a given concurrent system and it is, in fact, an information system, whereas the second one represents a transition relation between global states. It is assumed that both data tables include only part of possible observations (global states and transitions between them). In other words, they contain partial knowledge about the system behaviour. This approach is called the Open World Assumption. The knowledge included in data tables is represented in the form of rules extracted from them. Nevertheless, such partial knowledge is sufficient to construct a suitable model. The remaining knowledge (or - still - a part of it) can be discovered on the basis of the created models. The new knowledge derived from models can concern either the new states of systems or the new transitions between them which have not earlier been observed. It is worth noting that the new knowledge is not at variance with the possessed information about the system behaviour. The new knowledge about the system behaviour is extracted on the basis of occurrence graphs generated for the constructed net models. Each information system can be decomposed into subtables called the components of this system. Components define, in certain sense, the strongest functional modules of the system. Decomposition is a division of a given system into smaller, relatively independent subsystems. In ROSECON, the net models can also be built on the basis of decomposed information systems. In all the foregoing cases, the net construction consists of two stages. In the first stage, all dependencies presented as the deterministic rules in a modelled system are extracted from data tables. In the second stage, a net model is built. One of the important aspects of data mining processes is also the analysis of data changes in time. Data changing in time are called temporal (timed-dependent) data. In this case, the result of observation of the behaviour of a given concurrent system is shown in the form of a data table representing a temporal information system. Temporal information systems provide the knowledge about the behaviour of described systems in time. Rows (objects, in the information system terminology) in such systems are ordered in time. Each new row includes a global state of a system observed in the next time unit. A lot of modelled systems change their properties with time. Models built for given periods of observation must be reconstructed for the next periods in order to take new properties into consideration. The approach implemented in ROSECON is based on observation of the behaviour of a system in the so-called time windows being sets of global states of a system observed in individual periods of time. Attention is focused mainly on the functional dependencies between states of processes in a given concurrent system. It is observed how the functional dependencies between states of processes change in the consecutive periods of time. On the basis of observation, future model changes are predicted, because the model of a concurrent system is built using information about functional dependencies between states of processes. In order to extract the knowledge about model changes, ROSECON uses methods based on the rough set theory. This knowledge is needed to predict further model changes. The knowledge extracted from the possessed observation is represented by the so-called prediction matrix or by a temporal flow graph. It has an uncertain character, so prediction is characterized by some coefficients determining its degree of certainty. These coefficients are computed on the basis of observation which have been collected so far. Prediction matrices and temporal flow graphs provide some kind of prediction rules. To compute reducts and rules and to decompose information systems, ROSECON uses the standard methods based on the rough set theory. The concurrent models created by ROSECON have the form of coloured Petri nets. These nets are high-level nets introduced by Kurt Jensen. Models in the form of coloured Petri nets are coherent, readable, and their construction is relatively simple. The complexity of models is divided between net structures, declarations and net inscriptions. However, large data tables cause the models become more complicated. Especially, their declarations and inscriptions are less readable because of their sizes. A natural way to solve this problem is building hierarchical models. A given hierarchical model can be constructed by suitable combining a number of smaller models. ROSECON can be used to model concurrent systems described by information systems using some kind of hierarchical coloured Petri nets. First, the so-called generalized information systems are built. Such systems describe concurrent systems on the more abstract level by combining specific (real) processes of concurrent systems into some generalized processes. On the basis of generalized information systems, the net models on the first level are built. Subnets on the second level represent real processes.
Input data tables in ROSECON can describe concurrent systems understood widely, i.e., not only as computer systems (software or hardware). The concurrent system can be a system, for example, economic, meteorological, biological, genetic, etc., consisting of some separated processes whose local states are partly independent on the local states of other processes. ROSECON can be applied for modelling and analysing such systems as well as predicting model changes of concurrent systems in time.
Each input and output data file in ROSECON is written in the suitable XML format. The Extensible Markup Language (XML) allows the specification of specialized markup languages for the convenient exchange of information. It facilitates the interchange of data between different applications. ROSECON makes cooperation with the Design/CPN system, CPN Tools and CPNetwork possible. Design/CPN, CPN Tools (both created by Kurt Jensen's group at the Aarhus University) and CPNetwork support users with modelling and simulation with coloured Petri nets. All net models generated by ROSECON can be exported to the XML formats accepted by Design/CPN, CPN Tools and CPNetwork.
ROSECON is a software system for automatized synthesis of concurrent processes discovered in data tables. ROSECON currently offers the following main functions:
ROSECON is evolving continuously. The main tasks which remain to be implemented soon in ROSECON are the following: reconstruction of net models when description of concurrent systems changes in time and creating hierarchical net models for concurrent systems. Moreover, computing reducts and rules is equivalent to computing prime implicants of a Boolean function, i.e., it is an NP-hard problem. An exhaustive search is thus not suitable for large tables. Therefore, it is planned to implement some approximate methods in ROSECON.
Development of the ROSECON system has been partially supported by the grants Nos. 3 T11C 005 28 and 3 T11C 012 28 from Ministry of Scientific Research and Information Technology of the Republic of Poland.