Plenary Lecture

Classification Methods for Bibliomining

Prof. Ioana Moisil
Department of Computer Science and Automatic Control,
Hermann Oberth Faculty of Engineering
Lucian Blaga University of Sibiu
Blvd. Victoriei 10
550024 Sibiu
ROMANIA

Email: ioana.moisil@ulbsibiu.ro
Website: http://csac.ulbsibiu.ro

Abstract: Advances in information technology are having an important impact on library systems. Large collections of heterogeneous data, from ancient manuscripts to sounds, videos and spatial data are now available in electronic format. Digital libraries are capturing human knowledge and distributing it over the web. The increasing volume of data in today digital repositories and library data warehouses has determined a wide use of computer-based sophisticated analysis techniques. Special operation of data mining can be performed in order to answer questions of librarians and researchers in information science. In 2003, S. Nicholson and J. Stanton introduced a new term – bibliomining - for data mining library systems. Therefore bibliomining is a large umbrella incorporating all data mining methods based on mathematics, statistics, operational research, machine learning, evolutionary computing, visualization techniques, and including traditional methods of analysing groups of bibliographic references as authorship, publications, and literature, specific to bibliometrics. Librarians and researchers in information science are mining library data warehouses and other library data collections in order to discover patterns and to understand library users’ behaviour, their information and services needs, but also in order to evaluate and predict the effectiveness of library services, to discover trends in queries and to identify hot topics. Classification of items based on their characteristics (features, attributes, properties) in pre-defined categories is one of the most important bibliomining tasks. Classification is defined as the ordering of items in pre-defined groups (categories) or classes, based on their similarity. The classification process consists in assigning one of k labels (or classes) to each of n items derived from a specific problem. Classification predicts categorical labels. Analysis goal is to find a classification, a model or profile for each class that optimizes a combinatorial function consisting of assignment costs, based on the individual choice of label made for each item, and separation costs - based on the pair of choices made for two related items. In machine learning classification is defined as supervised learning. Classification, as a bibliomining technique, can be used for finding hidden patterns in data by deciding to what pre-defined class to assign a record of the data set, and also in prediction, to predict group membership for data instances. This lecture describes the most important classification methods (traditional approaches as classification trees, discriminant analysis, generalized linear models, modern statistical machine learning algorithms, support vector machine, belief networks, Gaussian processes, neural network, evolutionary algorithms, swarm intelligence, boosting and ensemble) and their use in mining library data collections. Research questions regarding pre-processing operations, attribute relevance and classifiers’ performance will also be discussed with emphasis on the specificity of the library items to be classified.

Brief Biography of the Speaker:
Ioana Moisil received the M.Sc. in Mathematics at the University of Bucharest, in 1971, the scientific grade in Statistical, Epidemiological and Operation Research Methods Applied in Public Health and Medicine at the Universite Libre de Bruxelles, in Belgium, in 1991 and the Ph.D. in Mathematics at the Romanian Academy in 1997. Work places: the National Institute for Research & Development in Informatics - I.C.I (1971-1986), Carol Davila Faculty of Medicine Bucharest – department of Biophysics, CCSSDM Center of the Ministry of Health. At present she is a full-time Professor and a Senior Researcher at the Department of Computer Science and Automatic Control – Faculty of Engineering at the “Lucian Blaga” University of Sibiu. She is the author/co-author of fourteen books and over 150 scientific papers. Her scientific interests include intelligent systems, healthcare telematics, web technologies, data-mining, e-learning, modelling and simulation, uncertainty management, human-computer interaction. Professor Moisil participated in several EU funded projects as project manager for the national partner (Telenurse ID ENTITY, MGT, PROPRACTITION, PRO-ACCESS), in Tempus projects and in national funded projects as research manager and software development coordinator (INFOSOC – eUNIV, AMTRANS – eCASTOR, INFOSOC - e-Scribe, INFOSOC – DANTE, e-EDU-Quality, eTransMobility, CNCSIS 2007-code 33, Studies on multivariate interpolation, polinomial classifiers and applications, CNCSIS 2007 – cod 1502, Aspects concerning the psycho-cognitive abilities of artificial intelligent agents and applications in ITC based education). Ioana Moisil is a member of EARLI (European Association for Research in Learning and Instruction), she is Romanian representative in the IMIA SIG and EFMI WG5 Nursing Informatics, honorary member of the Bohemian Medical Association J.E.Purkyne of Bio-engineering and Medical Informatics, member of the ISCB – International Society for Clinical Biostatistics – Romanian National Group, of the Romanian Association of Engineers, member of the IITM- International Institute of Tele-Medicine and of the Romanian Society of Mathematics Sciences. She is vice-president of the Romanian Medical Informatics Society; vice-president of the HIT Foundation for Health Informatics and Telematics and a member of RoCHI-ACM. Professor Moisil is taking part in several international peer-review committees and conferences scientific boards.