MACHINE LEARNING ALGORITHMS AND THEIR APPLICATION

->hrvatski

The project is supported by Croatian Ministry of Science Education and Sports
contract number: 098-0982560-2563
starting date: January 1st 2007
part of the program: Computational Knowledge Discovery in Scientific Applications

International Workshop e-Cardiology and European
Society of Cardiology working group meeting
organizer: dr. G.Krstačić

Handbook: Data Mining for Knowledge Discovery (in Croatian)
by D.Gamberger

Book: Foundations of Rule Learnig
by J.Fuernkranz, D.Gamberger, and N.Lavrac

Comparison of two methods for protein function annotation

Example of medical plan for the
case of pulmanory oedema

Decision support based on the
ontological knowledge representation

Procedural knowledge integrated into OWL ontology

Project summary

Effective knowledge handling is the limiting factor of computational intelligence. And although the project purpose is practical implementation of knowledge technology tasks, our main research topic is actually machine learning. Our previous work gives us good reasons to believe that machine learning algorithms are not only powerful tools for intelligent data analysis and knowledge discovery tasks, but also that they can help us to structure existing expert knowledge and that they can be the driving force for decision support procedures.

The topic is both theoretical and practical research related to machine learning algorithms. Special attention is devoted to feature construction in general, and specifically for inductive learning from different complex data forms including temporal signals, two-dimensional images, text, and relational databases. Theoretically and practically we try to prove usefulness of the saturation-based concept of inductive learning. We will work on noise handling and overfitting prevention techniques.

The goal is development of algorithms that can be effectively used in intelligent data analysis and knowledge handling tasks. Applications are in very different fields including but not restricted to medicine, chemistry, biology, and social sciences. In each of these domains, we strongly cooperate with respective domain experts trying to obtain novel results that cannot be obtained by other methods. The aim of this interdisciplinary work is to achieve results that are significant for the development of the target domain as well as for the computer science as an illustration of the quality and significance of applied algorithms.

Among others, the importance of the proposed project is in the fact that it presents necessary support for EU projects in which we are involved. The first has been HEARTFAID in the period from February 2006 to April 2009. In cooperation with ten other European partners within three years we realized a platform of services able to improve medical-clinical management related to heart failure disease. Within this project, we used our machine learning expertise for knowledge discovery tasks and for the development of effective knowledge representation and decision support services. Although the project used and developed modern computer science techniques, it has been strongly an interdisciplinary work with the result potentially very relevant also for medicine.

Currently we work on EU FP7 project e-LICO: An e-Laboratory for Interdisciplinary Collaborative Research in Data Mining and Data-Intensive Sciences. In the frames of the project we cooperate with a few leading European centers in the field of machine learning, workflow planning, and application of recommender systems. Besides that we started the work on the FP7 project Forecasting Financial Crisis (FOC).

Members

Members on HEARTFAID project

Technical support

Activities

Published papers in year 2013

Published papers in year 2012

Published papers in year 2011

Published papers in year 2010

Published papers in year 2009

Published papers in year 2008

Published papers in year 2007