IE 483x/583x Knowledge Discovery and Data Mining

Course Information (Back)

Objective

Upon completing this course students will understand how data mining can be used together with data warehousing and other knowledge discovery technologies to create a competetive advantage for the enterprise.  In particular students will know how to implement common data mining techniques, including statistical methods, machine learning methods, and visualization techniques to extract patterns, trends, and other useful information from databases.

Description

This course focuses on algorithm techniques that can be used for data mining tasks such as classification, association rule mining, clustering, and numerical prediction.  This includes probabilistic and statistical methods, genetic algorithms and neural networks, visualization techniques, and mathematical programming.  We also place such data mining within the larger picture of knowledge discovery in databases and in particular its relationship with data warehousing.  We will consider numereous case studies from both manufacturing and service industries.

Readings

Primary textbook:

The only required text is Ian H. Witten and Eibe Frank, (2000), Data Mining: practical machine learning tools and techniques with Java implementations, Morgan Kaufmann Publishers, San Fransisco, CA.  This book emphasizes algorithms for model induction, that is, the actual data mining rather than data preperation and other steps in the knowledge discovery process.

Secondary texts and resources:

• Other general textbooks:
• For a more business oriented introduction to data mining, written primarily for data mining professionals, check out Michael J.A. Berry and Gordon Linoff (2000), Mastering Data Mining: the art and science of customer relationship managment, John Wiley & Sons, New York.
• Another very good introductory textbook, that takas a more database oriented approach, is Jiawei Han and Micheline Kamber, (2001), Data Mining: Concepts and Techniques, Morgan Kaufmann Publishers, San Fransisco, CA.
• Some algorithms that will be discussed in class are not covered by the textbook but can be found in Tom Mitchell (1997),  Machine Learning, McGraw-Hill, Boston.  This text discusses many of the same concepts from a more traditional artificial intelligence perspective, including artificial neural networks and genetic algorithms.
• Clustering:
• The algorithms discussed in class but not covered by the primary textbook are discussed in "Efficient and Effective Clustering Methods for Spatial Data Mining" by Ng and Han, and "BIRCH: An Efficient Data Clustering Method for Very Large Databases" by Zhang, Ramakrishnan, and Livny.  They are also discussed in the book by Han and Kamber (see above).
• Association rule discovery:
• Association rule mining was introduced in the paper "Mining Association Rules between Sets of Items in Large Databases" by Agrawal, Imielinski, and Swami.
• The FP-tree is introduced in "Mining Frequent Patterns without Candidate Generation" by Han, Pei, and Yin and the CLOSET algorithm is developed in "CLOSET: An Efficient Algorithm for Mining Frequent Closed Itemsets" by Pei, Han, and Mao.  A short overview of this line of research is given in "Mining Frequent Patterns by Pattern Growth:  Methodology and Implications" by Pan and Han.
• Mining with a taxonomy is introduced in "Mining Generalized Association Rules" by Srikant and Agrawal.
• Application of association rule mining to a recommender is discussed in "Efficient Adaptive-Support Association Rule Mining for Recommender Systems" by Lin, Alvaraz, and Ruiz and can be found here.
• Optimization and data mining:
• "A Genetic Algorithm Tutorial" by Whitley.  Applying genetic algorithms for attributes selection is discussed in "Genetic Algorithms as a Tool for Feature Selection in Machine Learning" by Vafaie and De Jong, and  "Feature Subset Selection Using A Genetic Algorithm" by Yang and Honavar.  Evolutionary search for attribute selection for clustering is introduced in "Feature Selection in Unsupervised Learning via Evolutionary Search" by Kim, Street, and Menczer.
• For a fairly theoretical discussion of mathematical programming in data mining, read "Optimization Methods in Massive Data Sets" by Bradley, Mangasarian, and Musicant.  Two surveys of support vector machines: "A Tutorial on Support Vector Machines for Pattern Recognition" by Burges and "A Tutorial on Support Vector Regression" by Smola and Scholkopf.  A much shorter (but sufficient) review is done in "Support Vector Machines:  Hype or Hallelujah?" by Bennett and Campell.
• The paper "Local Learning in Probabilistic Networks with Hidden Variables" by Russel, Binder, Koller, and Kanazawa discusses using gradient-descent algorithms to train Bayesian networks.
• Web mining:
• The book Mining the Web by Soumen Chakrabarti (Morgan Kaufmann, 2003) covers all of the web mining material discussed in class (and more), as well as much of the material on complex data types and text mining.
• Explore the web resources for more information!