• Chap 1: Introduction
• Chap 2: Association Rule Analysis
• Chap 3: Negative Association Rule
• Chap 4: Causality
• Chap 5: Causal Rule Analysis
• Chap 6: Very large Database
• Chap 7: Small Database
• Chap 8: Conclusion
• Reference

Some important papers:

• The first paper (Apriori) 1993
• Improved Algorithm 1994
• Mining sequential pattern 1995
• Large database 1995
• Rules to correlations 1997
• Refinement of rule 1998
• Model based mining 2004
• Closed itemsets 2005
• Clustering data stream
• General Survey 1
• General Survey 2
• Survey slides

Association rule mining concerns the extraction of meaningful rules or relations from a large database. I believe historically, this is the area of Database research. It evolves into Data mining research. There are usually two applications of this mining technology: mining the Web and mining a business database. There are also two types of problems to be solved: data mining and data clustering. I think we can learn the fundamentals from the handouts of Jeff Ullman's course web site at Stanford.

So far, we have only applied Association rule analysis to build a "filter" for mining simulation data. For example, our recent paper mentions the use of Association rule analysis to establish a good ordering in decision-diagram based Boolean learning. Our vision to apply Association rule mining in building the simulation data mining tool can be illustrated in this presentation of project overview.

It is still under investigation how to apply association rule analysis in solving other problems. For example, how to apply it in test data mining? I believe we need a better understanding of the field, knowing the limitation and assumption in each algorithm in order to formulate meaningful problems in our applications.

However, for most of test and verification applications, I do not think that Association rule analysis alone can solve the problem entirely. I see Association rule analysis as the front-end data analysis engine to filter and organize data so that most focused techniques like machine learning or statistical learning can be applied develop the learned models that are useful to guide us in the applications.