Titel

1
Center for Computing Technologies
2
Computer supported analysis and assessment in
production processes11/05/98

• Dr. Ubbo Visser
• TZI - Center for Computing Technologies
• Department of Mathematics and Computer Science
• University of Bremen, Germany
• visser_at_tzi.de
• http//www.tzi.de

3
Contents

• TZI - Center of Computing Technologies
• Motivation, aims, structure, developments
• Areas, RD topics
• Potential application of Mobile Assistant and a
  future example
• Accident management and networking
• Welding for shipbuilding
• Where are the problems?
• Information Warehousing and Data Mining
• Concepts, models, OLAP
• Knowledge discovery, methods
• New method for knowledge discovery
• Qualitative abstraction
• Conclusion
• Combination of MA, IWH and DM

4
1. TZI - Center for Computing Technologies
5
Motivation Computer Science Applications

• Clear trend towards application areas
• Engineering
• Business
• Geoscience
• Medicine
• Computer Science in Bremen
• building bridges between application areas in
  computer science
• interdisciplinary co-operations within the
  University (e.g. environmental protection,
  logistics)

6
The TZI Objectives

• Transfer of computing technologies to industry
  through projects with software companies
• Development of innovative application-oriented
  technologies
• Interdisciplinary co-operation projects within
  the University
• Solid mix of basic research and
  application-oriented projects for students and
  staff

7
The Center for Computing Technologies
Safe Systems
Image Processing
Human Factor Soft- ware / Information Management
Intelligent Systems
Prof. Friedrich Prof. Kubicek
Prof. Bormann
Prof. Herzog Prof. Wischnewski
Prof. Herzog
Dr. U. Visser T. Waschulzik
U. Haupt
Dr. H. Schlingloff
Science and Industry Board
TZI
Director Prof. Herzog Managing Director Dr.
Günter Sekretary V. Landau
8
TZI facts and figures

• 10 professors
• 90 researchers
• External funding in 1997 US 3.3 million
• 65 research grants and 35 contracts
• Joint projects with
• 40 companies
• 12 government agencies
• numerous research institutions

9
TZI RD topics

• Quality control with innovative methods in image
  processing
• Intelligent search in multimedia archives
• Development, formal verification and test of
  reliable systems
• Check of systems regarding reliability
• Check of human-machine-interaction
• Forming ergonomic software
• Support with the introduction and management of
  complex information systems
• Knowledge-based methods in medicine, the
  environmental areas and production planning
  process
• Technical basics of computer based communication
  and co-operation
• Technologies for the development and process of
  multimedia documents

10
2. Potential application of Mobile Assistant and
a future example
11
Potential applicationAccident
management,Container terminal Bremerhaven
Accident Management Center

• Damage control
• Video analysis
• Decision support
• Numerous applications on WC

12
Future example for shipbuildingQuality
control
13
Architecture
Shipbuilding
14
Architecture
Shipbuilding
Mobile Assistant
15
Architecture
Shipbuilding
Mobile Assistant
Add. sensors
Audio
16
Architecture
Shipbuilding
Mobile Assistant
Add. sensors
Sensors
Audio
Infra-red
17
Architecture
Shipbuilding
Mobile Assistant
Add. sensors
Sensors
Audio
Infra-red
3D-distance camera

18
Architecture
Server
Shipbuilding
Mobile Assistant
Add. sensors
Sensors
Audio
Infra-red
3D-distance camera

19
Architecture
Server Information Warehouse
Shipbuilding
Mobile Assistant
Add. sensors
Sensors
Audio
Infra-red
3D-distance camera

20
Architecture
Server Information Warehouse
Shipbuilding
Mobile Assistant
Add. sensors
Sensors
Audio
Infra-red
3D-distance camera

Intra/Internet
21
Applications
Information Warehouse

• Simple applications
• data reduction
• simple analysis
• additional information

• Long-term analysis
• data mining
• complex/demanding analysis
• quantitive qualitative data

22
Problems or Motivation?

• The use of additional sensors to collect data and
  control quality of work
• Sensors audio, video, infrared, 3D distance
  camera
• 3D distance camera is constructed by Daimler Benz
  Aerospace, Bremen, world novelty on the Hannover
  exhibition in April 1998
• Multiple sensors that add additional information,
  which is difficult to handle without intelligent
  tools
• Multiple sources causes heterogeneous data

Information Warehouse and Data Mining
23
3. Information Warehousing andData Mining
24
Necessity is the mother of invention (Dr.
Jiawei Han)

• Information/Data warehousing Integrating data
  from multiple sources into large warehouses and
  support for on-line analytical processing and
  business decision making
• Data mining (knowledge discovery in databases)
  Extraction of interesting and new knowledge
  (rules, regularities, patterns, constraints)
  from data in large databases
• Necessity Data explosion problem ---
  computerized data collection tools and mature
  database technology leads to tremendous amounts
  of data stored in databases
• We are drowning in data, creating data cemeteries
  but lacking knowledge

25
Data Warehousing

• A data warehouse is a subject-oriented,
  integrated, time-variant, and nonvolatile
  collection of data in support of managements
  decision-making process. --- W. H. Inmon
• A data warehouse is
• a decision support database that is maintained
  separately from the organizations operational
  databases
• a integration of data from multiple heterogeneous
  sources to support the continuing need for
  structured and /or ad-hoc queries, analytical
  reporting, and decision support
• Existing solutions
• IBM Visual Warehouse, DB2 OLAP Server, Business
  Objects
• Oracle Data Warehouse, Data-Mart-Suite
• Hewlett-Packard HP Open Warehouse
• ...

26
OLAP On-Line Analytical Processing

• A multidimensional, LOGICAL view of the data
• Interactive analysis of the data drill, pivot,
  slice_dice, filter
• Summarization and aggregations at dimension
  intersections
• Retrieval and display of data in 2-D or 3-D
  crosstabs, charts, and graphs, with easy pivoting
  of axes
• Analytical modeling deriving ratios, variance,
  etc. and involving measurements or numerical data
  across many dimensions
• Forecasting, trend analysis and statistical
  analysis
• Requirement Quick response to OLAP queries
• Tools available Business Objects (IBM),
  Data-Mart-Suite (Oracle), ...

27
Views of Data Mining Techniques

• the knowledge to be discovered
• the database to be mined
• the techniques to be adopted

28
Knowledge to be discovered

• characterization
• summarization, generalization and contrast data
  characteristics
• association
• rules like buys(x,car) --gt buys(x,insurance)
• classification
• classify data based on the values in an
  attribute, e.g. classify insurance policies by
  the amount of premium
• clustering
• cluster data to form new classes, e.g. cluster
  RNA patterns in biological databases
• trend
• deviation
• pattern analysis
• ...

29
Data Mining Methods

• Database-oriented, multiple mining functions IBM
  Intelligent Miner, SGI MineSet, DBMiner, etc.
• OLAP-based (data warehousing) Concept Ltd.
  Business Objects, Oracle Data-Mart,
  Informix-MetaCube, Redbricks, Essbase, etc.
• Machine learning AQ15, ID3, C5.0, INLEN, Cobweb,
  etc.
• Statistical approaches, e.g., KnowledgeSeeker,
  Bayesian, Explora, etc.
• Visualization approach VisDB(Keim, Kriegel, et
  al.1994)
• Neural network approach, e.g., 4thoughts (Cognos)
• Rule-Extraction from neural networks, e.g.
  RuleNeg (QUT/NRC, 1995)
• Rough sets, fuzzy sets Datalogic/R, 49er, etc.
• Knowledge representation reasoning e.g., IMACS
• Inductive logic programming Muggleton Raedt
  1994, etc.
• Deductive DB integration KnowlegeMiner (Shen et
  al.1996)
• New approach for qualitative abstraction
  (Boronowski, TZI, 1998)

30
Integration of Data Mining and Data Warehousing

• Data warehouse provides clean and integrated data
  for fruitful mining
• Data mining provides powerful tools for analysis
  of data stored in data warehouses
• OLAP can be viewed as data summarization and
  simple data mining
• Data mining provides more analysis tools, e.g.,
  association, classification, clustering,
  pattern-directed, and trend analysis

31
4. New TZI Data Mining approach for
KDDqualitative abstraction
32
Qualitative abstraction
Collected data
Preparation
Qualitative abstraction
F(t)
ltl2,decgt, ltl1..l2,decgt, ... lte1,incgt, lte2,stdgt,
...
F(t)
Decisiontree- induction
Rules
33
Qualitative abstraction

• Reduced representation of functions
• Creates symbolic description
• Example for one function

t1 l1 std
t1...t2 l1...l3 inc
t2 l3 std
t2...t3 l2...l3 dec
t3 l2 std
t3...t4 l2...l4 inc
t4 l4 std
t4...t5 l4...l1 dec
t5 l1 std
34
Qualitative abstractionsingle function

• Identification of interesting points
• Abstraction of the value scale
• Abstraction of the time scale
• Creation of symbolic description

t
35
Qualitative abstractionsingle function

• Identification of interesting points
• Abstraction of the value scale
• Abstraction of the time scale
• Creating of the symbolic description

inf
minf
t
36
Qualitative abstractionsingle function

• Identification of interesting points
• Abstraction of the value scale
• Abstraction of the time scale
• Creating of the symbolic description

inf
t
minf
37
Qualitative abstractionsingle function
Time Landmark Gradient
t1 l1 std

• Identification of interesting points
• Abstraction of the value scale
• Abstraction of the time scale
• Creating symbolic description

inf
l4
l3
l2
l1
minf
t1
t2
t4
t3
t5
38
Qualitative abstractionsingle function
Time Landmark Gradient
t1 l1 std
t1...t2 l1...l3 inc

• Identification of interesting points
• Abstraction of the value scale
• Abstraction of the time scale
• Creating symbolic description

inf
l4
l3
l2
l1
minf
t1
t2
t4
t3
t5
39
Qualitative abstractionsingle function
Time Landmark Gradient
t1 l1 std
t1...t2 l1...l3 inc
t2 l3 std

• Identification of interesting points
• Abstraction of the value scale
• Abstraction of the time scale
• Creating symbolic description

t2...t3 l2...l3 dec
t3 l2 std
t3...t4 l2...l4 inc
t4 l4 std
t4...t5 l4...l1 dec
t5 l1 std
inf
l4
l3
l2
l1
minf
t1
t2
t4
t3
t5
40
Collected data
Preparation
Qualitative abstraction
F(t)
ltl2,decgt, ltl1..l2,decgt, ... lte1,incgt, lte2,stdgt,
...
F(t)
Decisiontree- induction
Rules
41
From data collection to rules

• What is the goal of this approach?
• Implicit knowledge in the data should become
  explicit
• Connections in multiple time series should become
  visible automatically
• e.g. if the pH-value is below 7.5 and the
  O2-content is increasing than the C02- contents
  is decreasing
• The rules have to be compehensive for an expert
• Treatment of huge databases with high dimensions
  must be feasible

42
5. Conclusion
43
Conclusion

• Hands-free wearable computers have excellent
  commercial potential
• Problems data processing, analysis of large
  amounts of data coming from multiple sources
• Data Warehouse, OLAP tools and Data Mining
  techniques to analyse data
• The combination of HFWC with modern sensors and
  intelligent software solutions, e.g. methods to
  discover new knowledge, will optimize the
  incredible technique