Introduction to Knowledge Discovery Center Professor Kesheng Wang

1
Introduction to Knowledge Discovery Center
Professor Kesheng Wang

• Department of Production and Quality
  EngineeringNorwegian University of Science and
  TechnologyN-7491 Trondheim, Norway
• Tel. [phone], Fax [phone]E-mail
  kesheng.wang_at_ipk.ntnu.no

2
Contents

• Purpose
• Personnel
• Resources
• Software
• Partners
• Current Research Areas
• Publications

3
Purpose
The Knowledge Discovery Center (KDC) conducts
research and education in data bases, data
warehouse, OLAP, data mining, knowledge
management, computational Intelligence (e.g.
Artificial Neural Networks, Fuzzy Logic Systems,
Genetic Algorithms), and other computational
intelligence theories for applications in
enterprises, such as business, industry,
healthcare, and service organizations. Funding
from government agencies and industrial
corporations has led to the development of new
solutions in areas such as engineering design
(e.g., process analysis, design automation,
autonomous decision making, and reengineering),
manufacturing (e.g., intelligent manufacturing,
system design, planning and scheduling,
reconfigurable systems, e-business, and system
maintenance and diagnosis), and medicine (e.g.,
disease diagnosis and prognosis, discovery of
medical knowledge, and healthcare business).
4
Personnel

• Kesheng Wang, Professor
• Morten Westvik, Researcher, SINTEF (Data Mining)
• Ove Rustung Hjlmvik, Researcher (Knowledge
  Management)
• Gelgele Hirpa, Associate Professor (Hybrid
  Intelligent System)
• Qiengfeng Yuan, Professor, SHU (Intelligent
  Robotics)
• Jianhui Li, PhD student (Intelligent Monitoring
  System)
• Paul Akangah, PhD student (Data Mining in Humain
  Related Science)
• Meng Tang, PhD Student (Hybrid Intelligent
  System)
• Ekateriva Pondmareva, PhD Student (Data Mining in
  Manufacturing)
• Sebastian Dransteld, PhD Student (ANN in
  Manufacturing)
• Hongzi Ma, PhD Student (Supply Chain Management)
• Dunderovic Ignor, MS Student (Data Mining in
  Healthcare)
• Jonathan Lienhardt, MS student (Gas for Robot
  calibration)
• Marc Reinhart, MS Student (Data mining for
  insurance policy)
• Aursand Morten, MS Student (Data mining for
  maintenance)
• Bård Røhne Halvorsen, MS student (Data Mining for
  healthcare)
• Some visiting researches from abroad

5
Resources

• Hardware
• Compaq Computers
• Laser printers
• Software
• Neuframe
• Genhenter
• IBM DB2
• IBM Intelligent Miner
• Clementine
• KnowledgePro expert system shells
• Xplain

6
Publicstions

• BOOKs (9)
• Computational Intelligence in Engineering
• Introduction to Knowledge Managemnet
• Intelligence condition monitoring and diagnosis
  systems
• .....
• Papers (more than 100)

7
Partners

• Statoil
• Hydro Aluminium
• Raufoss
• IBM scholar program
• SSPP
• SINTEF
• FEM Engineering
• Tenes AS
• Cognit AS
• SINOPEC, China
• SU, Shanghai
• IMC, UK
• .

8
Current Research Areas
The KDC pursues a dynamic research program that
reflects the progress of the industrial
engineering profession, as well as the needs of
the industrial and healthcare partners. Current
research include the following  topics

• Intelligent Manufacturing Systems
• Data Mining and Knowledge Discovery
• Computational Intelligence
• Enterprise Decision Making and Optimization
• Knowledge Management and Business Intelligence
  (product design, production planning and control,
  and production processes and systems)

9
Data mining in medical field

• Diagnosis for Heart Disease (Artifial Neural
  Networks approach)
• Breast Cancer Diagnosis (ANNs)
• Data mining for Healthcare business

10
Diagnosis for Heart Disease (Artifial Neural
Networks approach)
Data sets
11
Diagnosis for Heart Disease (Artifial Neural
Networks approach)
a age, s sex, p pain ( asympt, anang,
notang, angina), b blood pressure, c
cholesterol, f low fasting blood sugar, h
maximum heart rate, i induced angina, e
resting ECG (norm, abnorm, hyper), o
oldpeak, l slope (flat, up, down), t
thalamus (rev, norm, fixed)
12
Breast Cancer Diagnosis (ANNs)
Figure 2 Benign (left) and malignant (right)
cells
13
Breast Cancer Diagnosis (ANNs)

• Data sets (from University of Wisconsin
  hospitals)
• Take a FNA from the patients breasts
• Digital imagine from the FNA are transferred to a
  workstation by a video camera mounted on a
  microscope.
• These images are the input to the comuterized
  diagnostic system.
• The data consiss of 32 vector components. These
  vector components are based on the analysis of
  the three features area, texture and smoothness.
  

Network SBP RBF NEUfuzzy
Benign 98 88 88
Malign 97 84 87
14
Challenge of Health Care Business

• The challenge that faces most of health care
  organizations is that the volumes of data that
  can potentially be collected are so huge and the
  range of costumer behavior is so diverse that it
  seems impossible to rationalize what is
  happening.
• Data mining techniques can be used to help health
  care organization to make better decision quickly.

15
The Health Care Business Issues

• Can we identify indicators that are mainly
  responsible for the occurrence of special
  diseases like diabetes, thrombosis or
  tuberculosis?
• Which symptoms are highly correlated with
  positive examination tests?
• Can we set up a model that can predicate the
  patients stay in the hospital concerning a
  special disease?
• Can we detect medical indicators that act as an
  alarm system?
• Do the doctors who make the diagnosis observe the
  same treatment?

16
Some Student Projects

• How can we perform weight rating for Diagnosis
  Related Groups by using medical diagnosis?
• How can we perform patient profiling?
• Can we optimize medical prophylaxis tests?
• Can we detect pre-causes for a special medical
  condition?

17
Breast cancer diagnosis and treatment of patients
using data mining methods and techniques.
- It was a proposal to a research project and it
might be continuous to apply for strategy plan of
NTNU 2003-2008?
18
System architecture
19
Project development

• Phase 1 Pre-Screening
• It is deemed necessary to develop a database for
  pre-screening of patients.
• Phase 2 - Mammography
• Seek to develop diagnosis techniques and
  algorithms that will assist doctors in analysing
  MM-images.
• Phase 3 4 Fine Needle Aspiration Biopsy
• Seek to develop methodologies and diagnosis
  techniques and algorithms that will assist
  doctors in analysing FNA and Biopsies results
• Phase 5 Treatment
• Strive to achieve better and more efficient
  cancer treatment.

20
Project organisation

• Prof. Wang Keseng, Phd, NTNU
• Morten H. Westvik, Senior Advisor, IDM/SINTEF
• MD Steinar Lundgren, Phd, RiT/SINTEF UNIMED
• Lakhmi Jian, Phd, University of South Australia
• Yang Li, Phd, Western Michigan University
• Sweden Hospital
• Chinese Hospital