Home
Conference Officers
PC Committee
Conference Program
keynotes
research
short papers
Tutorials
Demostrations
panel
industrial
Registration
workshop
Venue
Accommodation

 

PANEL

P: Is there really anything beyond Frequent Patterns, Classification and Clustering in Data Mining?
Moderator: Vikram Pudi, IIIT Hyderabad, India
Abstract:Data mining, the automatic discovery of interesting patterns from large databases, has become a mainstream database research topic. Various kinds of patterns have been identified by researchers including classification and regression models, clusters, association rules, frequent patterns, time-series patterns, summaries, etc. Among these different data mining tasks, frequent pattern mining, classification and clustering have received most attention from the database research community. This is perhaps justified by the fact that most other data mining tasks can be reduced to these three. Here are some examples:
  • Association rules may be considered as merely another way to represent frequent itemsets.
  • Time-series are a special case of sequence databases and mining them reduces to mining frequent sequences.
  • Regression can be reduced to classification if the dependent attribute (i.e. the attribute whose values are to be estimated) is discretized into small ranges.
  • Outlier mining can be reduced to clustering, and then identifying data points that do not lie in any cluster.

In this context, the question naturally arises as to whether there is really anything more in data mining than frequent pattern mining, classification and clustering. This is an important question to be answered because if the question is answered in the affirmative, then future research may rather focus on the aspects of data mining that are beyond these basic tasks. On the other hand, if the question is answered in the negative, then we may well consider data mining as a "closed problem" because of the availability of very efficient algorithms for these three basic tasks.

The objective of this panel is to discuss this high impact and provocative question, which is basically a disguised way of asking: Do any fundamental research issues remain to be solved in data mining?