Jan Martinovič

Computational Intelligence Methods for Data Analysis and Mining of eLearning Activities

Enhancing the the effectiveness of web-based eduction has become one of the most important concerns within both educational engineering and information system fields. The development of information technologies has contributed to the growth in elearning as an important education method. This learning environment enables learners to participate in ’any time, any place’ personalized training. It has been known that the application of data mining and computational intelligent approaches can provide better learning environments, and in their effort to participate in this field, the authors introduced this study which consists in its first part of a survey of the applications of data mining and computational intelligence in web based education during (2004-2009), and the second part is a case study that aims to analyze students’ activities performed in a Learning Management System.

Analysis of Relations in eLearning

This paper is focused on searching for socials networks in eLearning system data. This data does not create a social network from the beginning, but contains records about student activity in an eLearning system. There are latent ties in these collections, which we also interpret as social network. We discovered these latent ties via clustering methods and, consequently, we used methods for social network analyses. The social networks we analyzed in this paper are represented by groups of students who have similar contacts and interact in similar social circles.

The ANTAREX approach to autotuning and adaptivity for energy efficient HPC systems

The ANTAREX project aims at expressing the application self-adaptivity through a Domain Specific Language (DSL) and to runtime manage and autotune applications for green and heterogeneous High Performance Computing (HPC) systems up to Exascale. The DSL approach allows the definition of energy-efficiency, performance, and adaptivity strategies as well as their enforcement at runtime through application autotuning and resource and power management. We show through a mini-app extracted from one of the project application use cases some initial exploration of application precision tuning by means enabled by the DSL.

A Tolerance Rough Set Based Overlapping Clustering for the DBLP Data

In the article there is presented comparison of overlapping clustering methods for data mining of DBLP datasets. For the analysis, the DBLP data sets were pre-processed, while each journal has been assigned attributes, defined by its topics. The data collection can be described as vague and uncertain; obtained clusters and applied queries do not necessarily have crisp boundaries. The authors presented clustering through a tolerance rough set method (TRSM) and fuzzy c-mean (FCM) algorithm for journal recommendation based on topic search. The comparison of both clustering methods was presented using different measures of similarity.In the information age, there are a lot of accounting information problems in the enterprises and government departments. At our current technological level, the paper sets up favorable accounting information sharing mechanism, including accounting information collection, delivery channels and so on, it can improve the efficiency of government departments and the management level through data sharing and computer intelligence processing. It will achieve efficiency economic management and higher overall efficiency and social benefits.

A Description of a Highly Modular System for the Emergent Flood Prediction

The main goal of our system is to provide the end user with information about an approaching disaster. The concept is to ensure information access to adequate data for all potential users, including citizens, local mayors, governments, and specialists, within one system. It is obvious that there is a knowledge gap between the lay user and specialist. Therefore, the system must be able to provide this information in a simple format for the less informed user while providing more complete information with computation adjustment and parameterization options to more qualified users. Important feature is the open structure and modular architecture that enables the usage of different modules. Modules can contain different functions, alternative simulations or additional features. Since the architectural structure is open, modules can be combined in any way to achieve any desired function in the system. One of many important modules is our own analytic solution to the flood waves for a small basin to our system.

Strategy Description for Mobile Embedded Control Systems Exploiting the Multi-agent Technology

Mobile embedded systems are a part of standard applications of distributed system control in real time. An example of a mobile control system is the robotic system. The software part of a distributed control system is realized by decision making and executive agents. The algorithm of agents cooperation was proposed with the control agent on a higher level. The algorithms for agents realized in robots are the same. Real---time dynamic simple strategy description and strategy learning possibility based on game observation is important for discovering opponents strategies and searching for tactical group movements, simulation and synthesis of suitable counter-strategies. For the improvement of game strategy, we are developing an abstract description of the game and propose ways to use this description (e.g. for learning rules and adapting team strategies to every single opponent).

Analysis of the DBLP Publication Classification Using Concept Lattices

The definitive classification of scientific journals depends on their aims and scopes details. In this paper, we present an approach to facilitate the journals classification of the DBLP datasets. For the analysis, the DBLP data sets were pre-processed by assigning each journal attributes defined by its topics and then the theory of formal concept analysis is introduced. It is subsequently shown how this theory can be applied to analyze the relations between journals and the extracted topics from their aims and scopes. The result is a concept lattice that contains information on journal-topic relational context depending on how they are associated. It is shown how this approach can be used to facilitate the classifications of scientific journals.

Robot Soccer - Strategy Description And Game Analysis.

The robot soccer game, as a part of standard applications of distributed system control in real time, provides numerous opportunities for the application of AI. Real-time dynamic strategy description and strategy learning possibility based on game observation are important to discover opponent’s strategies, search tactical group movements and synthesize proper counter-strategies. In this paper, the game is separated into physical part and logical part including strategy level and abstract level. Correspondingly, the game strategy description and prediction of ball motion are built up. The way to use this description, such as learning rules and adapting team strategies to every single opponent, is also discussed. Cluster analysis is used to validate the strategy extraction.

Control Loop Model of Virtual Company in BPM Simulation

The motivation of the paper is to introduce agent-based technology in the business process simulation. As in other cases, such simulation needs sufficient input data. However, in the case of business systems, real business data are not always available. Therefore, multi-agent systems often operate with randomly (resp. pseudo randomly) generated parameters. This method can also represent unpredictable phenomena. The core of the paper is to introduce the control loop model methodology in JADE business process simulation implementation. At the end of this paper the analysis of agent-based simulation outputs through process mining methods and methods for analysis of agents’ behavior in order to verify the correctness of used methodology is presented. The business process simulation inputs are randomly generated using the normal distribution. The results obtained show that using random number generation function with normal distribution can lead to the correct output data and therefore can be used to simulate real business processes.

Spectral Clustering: Left-Right-Oscillate Algorithm for Detecting Communities

Detection of communities in the complex networks is an actual problem solved in research area. The paper describes a new algorithm for this purpose. Left-Right-Oscillate algorithm (LRO) is based on spectral ordering of graph vertices. This approach allows us to detect a desired community – either by the size of the smallest communities or by the level of modularity. Since the LRO algorithm detects efficiently communities in large network even when these are not sharply partitioned, it turns to be specially suitable for the analysis of social, complex or coauthor networks. In this paper, proposed algorithm is used for finding communities in a large coauthor network - DBLP.