[Picture]

Knowledge Engineering and Data Mining

Knowledge Engineering and Discovery in Medical Domains

 I have been working in knowledge engineering and data mining since my MSc (1986/7). The MSc thesis concerned the application of machine learning to a logic database (a urology database from an English Hospital). Rather simplistically I consider data mining and knowledge discovery as part of a larger enterprise; that of knowledge engineering.

Current projects are:

  • Framework 7 Project: BRAVEHEALTH Patient Centric Approach for an Integrated, Adaptive, Context Aware Remote Diagnosis and Management of Cardiovascular Diseases
  • Phillips Healthcare Project: Advanced Medical Intelligence Predicting heart failure through application of data mining to large datasets.
  • Framework 7 Network of Excellence Project: Semantic Health SemanticHealthNet will develop a scalable and sustainable pan-European organisational and governance process for the semantic interoperability of clinical and biomedical knowledge, to help ensure that EHR systems are optimised for patient care, public health and clinical research across healthcare systems and institutions.
  • HEIF-5 University of Hull TeleHealth: Advancing Computational Frameworks for TeleHealth Two two year PDRAs investigating issues related to ongoing DCS Telehealth; in particular: Dependable and Adaptive Frameworks for TeleHealth; and, Computational Issues and Case Studies in TeleHealth.
  • SEED Funded PhD Project: Data Mining in Medicine Using Fuzzy Logic.Full DCS-SEED Scholarship to M. Mostafizur Rahman supervised by D.N. Davis.
  • University Funded PhD Project: Handling Complexities in Large Clinical Datasets.Full University Scholarship to Lisa Moore supervised by C.Kampbhampati, D.N. Davis and J. Cleland (HYMS).
  • KTP Project: E-Business Intelligence Improving business intelligence using data mining technology

    • Pattern Recognition Techniques used to date include:

    Tree Induction Over Logic Databases
    Rule Induction Over Logic Databases
    Statistical Classification for Image Feature Classification
    Evolvable Blackboard Architectures for Solution Optimisation
    MultiVariate Pattern Recognition for MicroFossil Classification
    Adaptive Contour Models for Neuroanatomical Feature Classification
    MultiLayer Perceptron and Support Vector Machines for CardioVascular Medicine
    Data Reduction using Pawlak Sets (Charlotte Bean)
    Tree Induction for CardioVascular Medicine using CART, DTREG, XpertRuleMiner, WEKA Algorithms, etc. etc.
    Association Rule Generation using WEKA Algorithms
    Bayesian Entropy Feature Selection and Filtering

    Current projects relate to the development of diagnostic models for CardioVascular Medicine with research students, clinicians at Hull Royal Infirmary and now an European Union Framework 7 Collaboration. This work makes use of Multi-Layer Perceptrons, Radial Basis Functions, Support Vector Machines, Self Organising Feature Maps and Information Theory for Attribute selection (e.g. Mutual Information, ReliefF) for Risk assessment in CardioVascular Medicine
    Fuzzy rule-based system applied to risk estimation of cardiovascular patients,
              Author Posting. (c) 'Copyright Holder', 2012. This is the author's version of the work. It is posted here by permission of 'Copyright Holder' for personal use, not for redistribution.   
       The definitive version is to be published in Journal of Multiple-Valued Logic and Soft Computing, 2012 (To Appear)
    Fuzzy Unordered Rules Induction Algorithm Used as Missing Value Imputation Methods for K-Mean Clustering on Real Cardiovascular Data
              ICDMKE_5 : The 2012 International Conference of Data Mining and Knowledge Engineering, World Congress on Engineering 2012 (WCE 2012) (July 4-6, 2012, London)
    Estimation of cardiovascular patient risk with a Bayesian network
              TRANSCOM 2011, 9th European Conference of Young Research and Scientific Workers, Žilina June 27 – 29, 2011 Slovak Republic
    Data mining applied to cardiovascular data
              Journal of Information Technologies Vol. 3, No. 2, November 2010, ISSN 1337-7469
    Generating and Verifying Risk Prediction Models Using Data Mining: A Case Study from Cardiovascular Medicine
              Published as Chapter in "Data Mining and Medical Knowledge Management: Cases and Applications,
              Editors: Petr Berka, Jan Rauch, & Djamel Abdelkader Zighed, IGI Global Inc. 2009.
    Predicting Cardiovascular Risks using Pattern Recognition and Data Mining
              Thuy Thi Thu Nguyen, Ph.D Thesis, Computer Science, University of Hull, August 2009.
    Generation and Verification of Risk Prediction Models for Carotid Endarterectomy using Data Mining and Neural Network Techniques
              European Society for Cardiovascular Surgery 57th Annual Congress of ESCVS, April 24-27, 2008 Barcelona Spain
    A Clustering Algorithm For Predicting CardioVascular Risk
              The 2007 International Conference of Data Mining and Knowledge Engineering, London, U.K., 2-4 July, 2007
    Feature Selection and Predicting CardioVascular Risk
              University of Hull Second Biosciences Workshop, December 2006.
    Predicting Cardiovascular Risks Using POSSUM, PPOSSUM and Neural Net Techniques
              ICEIS2006, 8th International Conference on Enterprise Information Systems, May 2006.
    Predicting CardioVascular Risk Using Neural Net Techniques
              University of Hull Biosciences Workshop, June 2005.
    The use of Artificial Neural Networks for risk prediction following Carotid endarterectomy
              Unpublished Paper, 2001.

    Multi-Agent Decision Support Systems (MADSS)
    Ongoing work in developing MADSS (Multiple Agent Decision Support System). MADSS is an agent-based decision support framework that can be of use in supplying decision-enabling information in a number of domains. This builds on my PhD and subsequent work in applying blackboard systems to medical image problems. We suggest that behaviours useful in solving problems, associated with specific information domains, results from designing specific architectures for particular types of agent communities. The grain of the design and architecture varies with the domain and task. Specific applications include machine vision in medicine, water supply infrastructure decision making, stock-trading portfolio management, cardio-vascular diagnosis and prognosis.
    Combining KADS with Zeus to Develop a Multi-Agent E-Commerce Application,
        International Journal of Electronic Commerce Research, 3(3-4):315-335, Kluwer, 2003
    A Multi-Agent System Framework for Decision Support in Stock Trading
       The IEEE Network Magazine Special Issue on Enterprise Networking and Services, Vol.16, No. 1, Jan/Feb 2002
    Information and Knowledge Exchange in a Multi-Agent system for Stock Trading
        IEEE/IEC Enterprise Networking Applications and Services Conference (EntNet2001), Atlanta, USA, July 2001.
    Using KADS to Design a Multi-Agent Framework for Stock Trading
      Agents for E-Business on the Internet, The 2001 International Multi-Conference Event, Las Vegas June 2001.
    Agent-Based Decision Support Framework for Water Supply Infrastructure and Development
        International Journal of Computers, Environment and Urban Systems, 24, 173-190, 2000
    A Multi-Agent Framework for Stock Trading
        World Computing Conference 2000, Beijing, August 2000
    Using KADS to Build Agents for E-Commerce
        Journal of Applied Software Systems, 2000
    The Application of Expert System and Agent Technology to Water Mains Rehabilitation Decision Making
        New Review of Applied Expert Systems, 5, 5-18, 1999
    An Agent Framework for Decision Support in the Water Industry
        International Conference on Artificial Intelligence Applied to Soft Computing, Honolulu. 1999.

    Tools used in building these systems include:

    weka
    CRoss Industry Standard Process for Data Mining (CRISP)
    Rosetta, CART etc
    Knowledge Analysis and Design System (KADS)
    XML as agent language
    KQML as agent language
    The Knowledge Acquisition Grid (KAG)
    File maintained by Dr D.N.Davis AT hull.ac.uk
    Last Updated December 2011