ROBERT KOPAL University College Algebra, Zagreb, Croatia
Social network analysis in national security
Social network analysis (SNA) provides an analytical/methodological framework for studying the patterns and processes that underpin financial transactions, phone calls, social/peer influences, information flow, local and international trade or any other exchange between individuals, groups, organizations or other entities. SNA can assist in detecting hidden connections between entities (modelled as network nodes) and reveal their degree of mutual influence at various levels, including the most complex global network level, which is can hardly be achieved with any other approach. In the context of growing concern of national security, SNA represents an evolution – if not a revolution – in unravelling the connectivity patterns among the security/intelligence agencies around the world. Looking at a sample of interesting cases, in this talk we will examine the range of SNA applications in national security.
MARKO BOHANEC Department of Knowledge Technologies, Jožef Stefan Institute, Slovenia
Data Mining for Decision Support and the “4C Requirements”: Correctness, Completeness, Consistency and Comprehensibility
With the abundance of data generated and stored in computer networks, the idea of developing decision support systems (DSSs) from data is becoming more and more appealing. Data can be viewed as a historical record of past decisions. The idea of the “Data Mining for Decision Support” approach is to process this data by machine learning and/or data mining algorithms in order to obtain models that can guide or predict future decisions. It is commonly believed that such models, once developed from data and checked for accuracy, are immediately suitable for supporting human decision makers and could be easily embedded in DSSs. Unfortunately, these expectations often turn out to be too naïve for practice and rarely work as expected. In this talk, we will investigate the “Data Mining for Decision Support” approach from the viewpoint of a decision analyst and DSS developer. We will use real cases from the areas of health-care management and food production to identify and illustrate the main obstacles for using data-mined models in DSSs. On this basis, we will formulate and explain the “4C Requirements” for models to be used in the DSSs: Correctness, Completeness, Consistency and Comprehensibility. For future data mining research, we will suggest to put more emphasis on considering ordered features and classes, fulfilling the dominance principle and monotonicity of models, measuring and improving the comprehensibility of models, and ensuring the completeness of models. Also, data mining and decision support should work together to achieve a better verification and validation of models with respect to the end-user’s problem, and provide a better support for integrating data with expert knowledge.
LUKA KRONEGGER Faculty of Social Sciences, University of Ljubljana, Slovenia
Bibliometric analysis of Slovenian scientific community
In recent years bibliographic analysis of research performance became a global hype. Physicists, Computer scientists, Organisational scientists, Social scientists, Librarians, etc. try to utilize given resources of bibliographic databases and computer power that enabled researchers to grasp these huge datasets. There are different ways how to analyse these data. Some see the bibliographic data as solid technical playground, others as fascinating quantification of researchers daily reality. The perspective of our interdisciplinary team tries to join these different views, by joining quantitative analysis with social theories and sensitivity to social reality. Several approaches to analyse publication activity of researchers in Slovenia, as small scientific community, will be presented. The presentation is mainly based on the results obtained by three methods: i) clustering of symbolic data applied on distributions of collaboration on the level of scientific disciplines, ii) modelling of network dynamics with SIENA, where models of preferential attachment and small world were tested, and iii) multilevel analysis with scientific productivity and excellence as dependent variables. The analyses are performed on co-authorship network of all Slovenian researchers (around 20000) affiliated into 72 scientific disciplines which are nested into 7 research fields. In the period 1986-2010 these researchers published around one million publications.
LOVRO ŠUBELJ Faculty of Computer and Information Science, University of Ljubljana, Slovenia
Reliability of bibliographic databases for scientometric network analysis
Bibliographic databases range from expensive hand-curated solutions (WoS) to preprint repositories (arXiv), public servers (DBLP) and automated services that collect manuscripts from the Web (Scholar). These provide the basis for scientific research, where new knowledge is derived from the existing, while also the main source of its evaluation. The databases are used by scientists on a daily basis and often studied in bibliometrics and scientometrics literature. However, while the content and structure of modern databases differ substantially, there exist only informal notions on their reliability. In this talk I will present a study of reliability of different citation and collaboration networks extracted from most popular bibliographic databases. Despite considerable differences between the databases, the results will indicate that there is no “best” database. The most appropriate choice largely depends on the type of the information one is interested in and on the type of the analysis one wishes to conduct.