THE INSTRUCTOR

Dr. Chia-Hui Chang - Rm E1-323, Ext: 4523
Homepage: http://www.csie.ncu.edu.tw/~chia/
Email: chia@csie.ncu.edu.tw

TEXTBOOK

• Data Mining:  Concepts and Techniques, J. Han and M. Kamber, Morgan Kaufmann , 2000
Slides can be downloaded from http://www.cs.sfu.ca/~han/DM_Book.html

REFERENCES

• Predictive Data Mining, S.M. Weiss and N. Indurkhya, Morgan Kaufmann, 1998
• Data Mining: Practical Machine Learning Tools and Techniques with Java Implementation, Ian H. Witten and Eibe Frank, Morgan Kaufman, 1999
• Advances in Knowledge Discovery and Data Mining, U.M. Fayyad, G. Piatetsky-Shapiro, P. Smyth and R. Uthurusam, AAAI/MIT Press, 1996:  A recent collection of research papers
• Mastering Data Mining: The Art and Science of Customer Relationship Management, by Michael J. A. Berry and Gordon Linoff, John Wiley & Sons; 1999

SCHEDULE & GRADING

 

Schedule
Introduction	week 2
Data Warehouse and OLAP	week 3
Data Preparation	week 4
Assignment 1 (MSSQL OLAP) -- Due. Mar. 19	week 4
DMQL & DBMiner Presentation Proposal -- Due Mar. 26	week 5
Characterization and Discrimination	week 6
Assignment 2 (DBMiner) -- Due. Apr. 9	week 7
Association Rule	week 8
Assignment 3 (Association Rule) -- Due. Apr. 30 (postponed to May. 7)	week 9
Classification	week 10
Clustering	week 11
Case Study	week 12
Exam, May. 14 -- 30%	week 13
Term Project -- 30%	3 weeks
Others -- 10% New [Selected or Voluntary] Assignment presentation [Optional] Oral presentation topics: Statistics: multiple variable analysis, Multi-dimentional discretization Association Rules: quantitative association rules, visualization Classification: bayesian belief networks, neural network, fuzzy set Clustering: density-based, grid-based, model-based methods, etc. Bibliographic notes in each chapter: Chapter 3: attribute set selection, multidimensional index trees, automatic discretization Chpater 2: discovery-oriented data cubes

COURSE CONTENT

1. Data Mining Techniques
• Associations -- finding association rules, dependency/belief networks, market basket analysis
• Classification -- building a classification model  (Decision tree | Rules | Neural network | Bayesian | Other )
• Clustering - for finding clusters or segments
• Sequential Patterns - tools for finding sequential patterns
2. Data Mining Software -- Libraries and Developer Kits for creating embedded data mining applications
• DBMiner
• IBM Intelligent Miner for Data
3. Data Mining Process
• Data Preparation: cleaning and transformation
• Choose type of data mining
• Interpret results
4. Web Mining, clickstream and session log analysis

TERM PROJECT

You have two options for the term project in this course: implementation-based and application-based. You can, based on your preference, choose either one. They have the same weight (30%) in your final grade.

Option I: Implmentation-based Project

There are three typical kinds of knowledge in data mining. They can be described as:

• Classification
• Association
• Clustering

You are required to choose any one of these algorithms and implement it using C++ or C, and test your program using some real data which can be obtained  from University of California-Irvine: Machine Learning Database Repositories. Please spend some time navigating through the different data sets there and  select the most suitable one for your testing. You might select several others to test your program to show your program does work.

You are required to prepare a documentation of your project, including the description of your project, the algorithm, design diagram, key data structures, source code, and  the testing results (input/output).  You need to explain your test and test results, including any references to help people understand the significance and interestingness of your work.

Option II: Application-based Project

The students are asked to choose one application domain, and prepare the documentation for your case study including:

1. The application case.
2. How do you prepare for your data?
3. Choose the mining type
4. How would you explain your result?
5. What problems you might encounter?

Note: The documentation should be printed using in 12pt font, single line spacing, and should not exceed 15 pages. Please also prepare 30-minute slides to present your work. The length of the essay, though not strictly required, should be between 10 to 15 pages. However, we pay more attention to the quality of your essay, not just the number of pages. Test data can be download from Datasets for Machine Learning, Knowledge Discovery and Data Mining.

CONFERENCES

• ACM KDD

http://informatik.uni-trier.de/~ley/db/conf/kdd/index.html
• IEEE ICDE
  http://informatik.uni-trier.de/~ley/db/conf/icde/index.html
• IEEE ICDM
  http://informatik.uni-trier.de/~ley/db/conf/icdm/index.html
• SIAM International Conference on Data Mining
  SDM03, SDM02, SDM01
• MLDM
  http://informatik.uni-trier.de/~ley/db/conf/mldm/index.html
• DaWaK
  http://informatik.uni-trier.de/~ley/db/conf/dawak/index.html
• DMKD
  http://informatik.uni-trier.de/~ley/db/conf/dmkd/index.html
• PKDD
  http://informatik.uni-trier.de/~ley/db/conf/pkdd/index.html
• PAKDD

htt~p://informatik.uni-trier.de/~ley/db/conf/pakdd/index.html

JOURNALS

• Data Mining and Konwledge Discovery

http://www.wkap.nl/journalhome.htm/1384-5810

REFERENCES

• KDNuggets Directory: Data Mining and Knowledge Discovery

http://www.kdnuggets.com/index.html
• Data Mining Software

http://www.cs.bham.ac.uk/~anp/software.html